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 #include <linux/selinux.h>
  62
  63 #include "audit.h"
  64
  65 extern struct list_head audit_filter_list[];
  66
  67 /* No syscall auditing will take place unless audit_enabled != 0. */
  68 extern int audit_enabled;
  69
  70 /* AUDIT_NAMES is the number of slots we reserve in the audit_context
  71  * for saving names from getname(). */
  72 #define AUDIT_NAMES    20
  73
  74 /* AUDIT_NAMES_RESERVED is the number of slots we reserve in the
  75  * audit_context from being used for nameless inodes from
  76  * path_lookup. */
  77 #define AUDIT_NAMES_RESERVED 7
  78
  79 /* When fs/namei.c:getname() is called, we store the pointer in name and
  80  * we don't let putname() free it (instead we free all of the saved
  81  * pointers at syscall exit time).
  82  *
  83  * Further, in fs/namei.c:path_lookup() we store the inode and device. */
  84 struct audit_names {
  85         const char      *name;
  86         unsigned long   ino;
  87         unsigned long   pino;
  88         dev_t           dev;
  89         umode_t         mode;
  90         uid_t           uid;
  91         gid_t           gid;
  92         dev_t           rdev;
  93         u32             osid;
  94 };
  95
  96 struct audit_aux_data {
  97         struct audit_aux_data   *next;
  98         int                     type;
  99 };
 100
 101 #define AUDIT_AUX_IPCPERM       0
 102
 103 struct audit_aux_data_ipcctl {
 104         struct audit_aux_data   d;
 105         struct ipc_perm         p;
 106         unsigned long           qbytes;
 107         uid_t                   uid;
 108         gid_t                   gid;
 109         mode_t                  mode;
 110         u32                     osid;
 111 };
 112
 113 struct audit_aux_data_socketcall {
 114         struct audit_aux_data   d;
 115         int                     nargs;
 116         unsigned long           args[0];
 117 };
 118
 119 struct audit_aux_data_sockaddr {
 120         struct audit_aux_data   d;
 121         int                     len;
 122         char                    a[0];
 123 };
 124
 125 struct audit_aux_data_path {
 126         struct audit_aux_data   d;
 127         struct dentry           *dentry;
 128         struct vfsmount         *mnt;
 129 };
 130
 131 /* The per-task audit context. */
 132 struct audit_context {
 133         int                 in_syscall; /* 1 if task is in a syscall */
 134         enum audit_state    state;
 135         unsigned int        serial;     /* serial number for record */
 136         struct timespec     ctime;      /* time of syscall entry */
 137         uid_t               loginuid;   /* login uid (identity) */
 138         int                 major;      /* syscall number */
 139         unsigned long       argv[4];    /* syscall arguments */
 140         int                 return_valid; /* return code is valid */
 141         long                return_code;/* syscall return code */
 142         int                 auditable;  /* 1 if record should be written */
 143         int                 name_count;
 144         struct audit_names  names[AUDIT_NAMES];
 145         struct dentry *     pwd;
 146         struct vfsmount *   pwdmnt;
 147         struct audit_context *previous; /* For nested syscalls */
 148         struct audit_aux_data *aux;
 149
 150                                 /* Save things to print about task_struct */
 151         pid_t               pid;
 152         uid_t               uid, euid, suid, fsuid;
 153         gid_t               gid, egid, sgid, fsgid;
 154         unsigned long       personality;
 155         int                 arch;
 156
 157 #if AUDIT_DEBUG
 158         int                 put_count;
 159         int                 ino_count;
 160 #endif
 161 };
 162
 163
 164 /* Compare a task_struct with an audit_rule.  Return 1 on match, 0
 165  * otherwise. */
 166 static int audit_filter_rules(struct task_struct *tsk,
 167                               struct audit_krule *rule,
 168                               struct audit_context *ctx,
 169                               enum audit_state *state)
 170 {
 171         int i, j, need_sid = 1;
 172         u32 sid;
 173
 174         for (i = 0; i < rule->field_count; i++) {
 175                 struct audit_field *f = &rule->fields[i];
 176                 int result = 0;
 177
 178                 switch (f->type) {
 179                 case AUDIT_PID:
 180                         result = audit_comparator(tsk->pid, f->op, f->val);
 181                         break;
 182                 case AUDIT_UID:
 183                         result = audit_comparator(tsk->uid, f->op, f->val);
 184                         break;
 185                 case AUDIT_EUID:
 186                         result = audit_comparator(tsk->euid, f->op, f->val);
 187                         break;
 188                 case AUDIT_SUID:
 189                         result = audit_comparator(tsk->suid, f->op, f->val);
 190                         break;
 191                 case AUDIT_FSUID:
 192                         result = audit_comparator(tsk->fsuid, f->op, f->val);
 193                         break;
 194                 case AUDIT_GID:
 195                         result = audit_comparator(tsk->gid, f->op, f->val);
 196                         break;
 197                 case AUDIT_EGID:
 198                         result = audit_comparator(tsk->egid, f->op, f->val);
 199                         break;
 200                 case AUDIT_SGID:
 201                         result = audit_comparator(tsk->sgid, f->op, f->val);
 202                         break;
 203                 case AUDIT_FSGID:
 204                         result = audit_comparator(tsk->fsgid, f->op, f->val);
 205                         break;
 206                 case AUDIT_PERS:
 207                         result = audit_comparator(tsk->personality, f->op, f->val);
 208                         break;
 209                 case AUDIT_ARCH:
 210                         if (ctx)
 211                                 result = audit_comparator(ctx->arch, f->op, f->val);
 212                         break;
 213
 214                 case AUDIT_EXIT:
 215                         if (ctx && ctx->return_valid)
 216                                 result = audit_comparator(ctx->return_code, f->op, f->val);
 217                         break;
 218                 case AUDIT_SUCCESS:
 219                         if (ctx && ctx->return_valid) {
 220                                 if (f->val)
 221                                         result = audit_comparator(ctx->return_valid, f->op, AUDITSC_SUCCESS);
 222                                 else
 223                                         result = audit_comparator(ctx->return_valid, f->op, AUDITSC_FAILURE);
 224                         }
 225                         break;
 226                 case AUDIT_DEVMAJOR:
 227                         if (ctx) {
 228                                 for (j = 0; j < ctx->name_count; j++) {
 229                                         if (audit_comparator(MAJOR(ctx->names[j].dev),  f->op, f->val)) {
 230                                                 ++result;
 231                                                 break;
 232                                         }
 233                                 }
 234                         }
 235                         break;
 236                 case AUDIT_DEVMINOR:
 237                         if (ctx) {
 238                                 for (j = 0; j < ctx->name_count; j++) {
 239                                         if (audit_comparator(MINOR(ctx->names[j].dev), f->op, f->val)) {
 240                                                 ++result;
 241                                                 break;
 242                                         }
 243                                 }
 244                         }
 245                         break;
 246                 case AUDIT_INODE:
 247                         if (ctx) {
 248                                 for (j = 0; j < ctx->name_count; j++) {
 249                                         if (audit_comparator(ctx->names[j].ino, f->op, f->val) ||
 250                                             audit_comparator(ctx->names[j].pino, f->op, f->val)) {
 251                                                 ++result;
 252                                                 break;
 253                                         }
 254                                 }
 255                         }
 256                         break;
 257                 case AUDIT_LOGINUID:
 258                         result = 0;
 259                         if (ctx)
 260                                 result = audit_comparator(ctx->loginuid, f->op, f->val);
 261                         break;
 262                 case AUDIT_SE_USER:
 263                 case AUDIT_SE_ROLE:
 264                 case AUDIT_SE_TYPE:
 265                 case AUDIT_SE_SEN:
 266                 case AUDIT_SE_CLR:
 267                         /* NOTE: this may return negative values indicating
 268                            a temporary error.  We simply treat this as a
 269                            match for now to avoid losing information that
 270                            may be wanted.   An error message will also be
 271                            logged upon error */
 272                         if (f->se_rule) {
 273                                 if (need_sid) {
 274                                         selinux_task_ctxid(tsk, &sid);
 275                                         need_sid = 0;
 276                                 }
 277                                 result = selinux_audit_rule_match(sid, f->type,
 278                                                                   f->op,
 279                                                                   f->se_rule,
 280                                                                   ctx);
 281                         }
 282                         break;
 283                 case AUDIT_ARG0:
 284                 case AUDIT_ARG1:
 285                 case AUDIT_ARG2:
 286                 case AUDIT_ARG3:
 287                         if (ctx)
 288                                 result = audit_comparator(ctx->argv[f->type-AUDIT_ARG0], f->op, f->val);
 289                         break;
 290                 }
 291
 292                 if (!result)
 293                         return 0;
 294         }
 295         switch (rule->action) {
 296         case AUDIT_NEVER:    *state = AUDIT_DISABLED;       break;
 297         case AUDIT_POSSIBLE: *state = AUDIT_BUILD_CONTEXT;  break;
 298         case AUDIT_ALWAYS:   *state = AUDIT_RECORD_CONTEXT; break;
 299         }
 300         return 1;
 301 }
 302
 303 /* At process creation time, we can determine if system-call auditing is
 304  * completely disabled for this task.  Since we only have the task
 305  * structure at this point, we can only check uid and gid.
 306  */
 307 static enum audit_state audit_filter_task(struct task_struct *tsk)
 308 {
 309         struct audit_entry *e;
 310         enum audit_state   state;
 311
 312         rcu_read_lock();
 313         list_for_each_entry_rcu(e, &audit_filter_list[AUDIT_FILTER_TASK], list) {
 314                 if (audit_filter_rules(tsk, &e->rule, NULL, &state)) {
 315                         rcu_read_unlock();
 316                         return state;
 317                 }
 318         }
 319         rcu_read_unlock();
 320         return AUDIT_BUILD_CONTEXT;
 321 }
 322
 323 /* At syscall entry and exit time, this filter is called if the
 324  * audit_state is not low enough that auditing cannot take place, but is
 325  * also not high enough that we already know we have to write an audit
 326  * record (i.e., the state is AUDIT_SETUP_CONTEXT or AUDIT_BUILD_CONTEXT).
 327  */
 328 static enum audit_state audit_filter_syscall(struct task_struct *tsk,
 329                                              struct audit_context *ctx,
 330                                              struct list_head *list)
 331 {
 332         struct audit_entry *e;
 333         enum audit_state state;
 334
 335         if (audit_pid && tsk->tgid == audit_pid)
 336                 return AUDIT_DISABLED;
 337
 338         rcu_read_lock();
 339         if (!list_empty(list)) {
 340                 int word = AUDIT_WORD(ctx->major);
 341                 int bit  = AUDIT_BIT(ctx->major);
 342
 343                 list_for_each_entry_rcu(e, list, list) {
 344                         if ((e->rule.mask[word] & bit) == bit
 345                                         && audit_filter_rules(tsk, &e->rule, ctx, &state)) {
 346                                 rcu_read_unlock();
 347                                 return state;
 348                         }
 349                 }
 350         }
 351         rcu_read_unlock();
 352         return AUDIT_BUILD_CONTEXT;
 353 }
 354
 355 static inline struct audit_context *audit_get_context(struct task_struct *tsk,
 356                                                       int return_valid,
 357                                                       int return_code)
 358 {
 359         struct audit_context *context = tsk->audit_context;
 360
 361         if (likely(!context))
 362                 return NULL;
 363         context->return_valid = return_valid;
 364         context->return_code  = return_code;
 365
 366         if (context->in_syscall && !context->auditable) {
 367                 enum audit_state state;
 368                 state = audit_filter_syscall(tsk, context, &audit_filter_list[AUDIT_FILTER_EXIT]);
 369                 if (state == AUDIT_RECORD_CONTEXT)
 370                         context->auditable = 1;
 371         }
 372
 373         context->pid = tsk->pid;
 374         context->uid = tsk->uid;
 375         context->gid = tsk->gid;
 376         context->euid = tsk->euid;
 377         context->suid = tsk->suid;
 378         context->fsuid = tsk->fsuid;
 379         context->egid = tsk->egid;
 380         context->sgid = tsk->sgid;
 381         context->fsgid = tsk->fsgid;
 382         context->personality = tsk->personality;
 383         tsk->audit_context = NULL;
 384         return context;
 385 }
 386
 387 static inline void audit_free_names(struct audit_context *context)
 388 {
 389         int i;
 390
 391 #if AUDIT_DEBUG == 2
 392         if (context->auditable
 393             ||context->put_count + context->ino_count != context->name_count) {
 394                 printk(KERN_ERR "%s:%d(:%d): major=%d in_syscall=%d"
 395                        " name_count=%d put_count=%d"
 396                        " ino_count=%d [NOT freeing]\n",
 397                        __FILE__, __LINE__,
 398                        context->serial, context->major, context->in_syscall,
 399                        context->name_count, context->put_count,
 400                        context->ino_count);
 401                 for (i = 0; i < context->name_count; i++) {
 402                         printk(KERN_ERR "names[%d] = %p = %s\n", i,
 403                                context->names[i].name,
 404                                context->names[i].name ?: "(null)");
 405                 }
 406                 dump_stack();
 407                 return;
 408         }
 409 #endif
 410 #if AUDIT_DEBUG
 411         context->put_count  = 0;
 412         context->ino_count  = 0;
 413 #endif
 414
 415         for (i = 0; i < context->name_count; i++) {
 416                 if (context->names[i].name)
 417                         __putname(context->names[i].name);
 418         }
 419         context->name_count = 0;
 420         if (context->pwd)
 421                 dput(context->pwd);
 422         if (context->pwdmnt)
 423                 mntput(context->pwdmnt);
 424         context->pwd = NULL;
 425         context->pwdmnt = NULL;
 426 }
 427
 428 static inline void audit_free_aux(struct audit_context *context)
 429 {
 430         struct audit_aux_data *aux;
 431
 432         while ((aux = context->aux)) {
 433                 if (aux->type == AUDIT_AVC_PATH) {
 434                         struct audit_aux_data_path *axi = (void *)aux;
 435                         dput(axi->dentry);
 436                         mntput(axi->mnt);
 437                 }
 438
 439                 context->aux = aux->next;
 440                 kfree(aux);
 441         }
 442 }
 443
 444 static inline void audit_zero_context(struct audit_context *context,
 445                                       enum audit_state state)
 446 {
 447         uid_t loginuid = context->loginuid;
 448
 449         memset(context, 0, sizeof(*context));
 450         context->state      = state;
 451         context->loginuid   = loginuid;
 452 }
 453
 454 static inline struct audit_context *audit_alloc_context(enum audit_state state)
 455 {
 456         struct audit_context *context;
 457
 458         if (!(context = kmalloc(sizeof(*context), GFP_KERNEL)))
 459                 return NULL;
 460         audit_zero_context(context, state);
 461         return context;
 462 }
 463
 464 /**
 465  * audit_alloc - allocate an audit context block for a task
 466  * @tsk: task
 467  *
 468  * Filter on the task information and allocate a per-task audit context
 469  * if necessary.  Doing so turns on system call auditing for the
 470  * specified task.  This is called from copy_process, so no lock is
 471  * needed.
 472  */
 473 int audit_alloc(struct task_struct *tsk)
 474 {
 475         struct audit_context *context;
 476         enum audit_state     state;
 477
 478         if (likely(!audit_enabled))
 479                 return 0; /* Return if not auditing. */
 480
 481         state = audit_filter_task(tsk);
 482         if (likely(state == AUDIT_DISABLED))
 483                 return 0;
 484
 485         if (!(context = audit_alloc_context(state))) {
 486                 audit_log_lost("out of memory in audit_alloc");
 487                 return -ENOMEM;
 488         }
 489
 490                                 /* Preserve login uid */
 491         context->loginuid = -1;
 492         if (current->audit_context)
 493                 context->loginuid = current->audit_context->loginuid;
 494
 495         tsk->audit_context  = context;
 496         set_tsk_thread_flag(tsk, TIF_SYSCALL_AUDIT);
 497         return 0;
 498 }
 499
 500 static inline void audit_free_context(struct audit_context *context)
 501 {
 502         struct audit_context *previous;
 503         int                  count = 0;
 504
 505         do {
 506                 previous = context->previous;
 507                 if (previous || (count &&  count < 10)) {
 508                         ++count;
 509                         printk(KERN_ERR "audit(:%d): major=%d name_count=%d:"
 510                                " freeing multiple contexts (%d)\n",
 511                                context->serial, context->major,
 512                                context->name_count, count);
 513                 }
 514                 audit_free_names(context);
 515                 audit_free_aux(context);
 516                 kfree(context);
 517                 context  = previous;
 518         } while (context);
 519         if (count >= 10)
 520                 printk(KERN_ERR "audit: freed %d contexts\n", count);
 521 }
 522
 523 static void audit_log_task_context(struct audit_buffer *ab)
 524 {
 525         char *ctx = NULL;
 526         ssize_t len = 0;
 527
 528         len = security_getprocattr(current, "current", NULL, 0);
 529         if (len < 0) {
 530                 if (len != -EINVAL)
 531                         goto error_path;
 532                 return;
 533         }
 534
 535         ctx = kmalloc(len, GFP_KERNEL);
 536         if (!ctx)
 537                 goto error_path;
 538
 539         len = security_getprocattr(current, "current", ctx, len);
 540         if (len < 0 )
 541                 goto error_path;
 542
 543         audit_log_format(ab, " subj=%s", ctx);
 544         return;
 545
 546 error_path:
 547         if (ctx)
 548                 kfree(ctx);
 549         audit_panic("error in audit_log_task_context");
 550         return;
 551 }
 552
 553 static void audit_log_task_info(struct audit_buffer *ab, struct task_struct *tsk)
 554 {
 555         char name[sizeof(tsk->comm)];
 556         struct mm_struct *mm = tsk->mm;
 557         struct vm_area_struct *vma;
 558
 559         /* tsk == current */
 560
 561         get_task_comm(name, tsk);
 562         audit_log_format(ab, " comm=");
 563         audit_log_untrustedstring(ab, name);
 564
 565         if (mm) {
 566                 down_read(&mm->mmap_sem);
 567                 vma = mm->mmap;
 568                 while (vma) {
 569                         if ((vma->vm_flags & VM_EXECUTABLE) &&
 570                             vma->vm_file) {
 571                                 audit_log_d_path(ab, "exe=",
 572                                                  vma->vm_file->f_dentry,
 573                                                  vma->vm_file->f_vfsmnt);
 574                                 break;
 575                         }
 576                         vma = vma->vm_next;
 577                 }
 578                 up_read(&mm->mmap_sem);
 579         }
 580         audit_log_task_context(ab);
 581 }
 582
 583 static void audit_log_exit(struct audit_context *context, struct task_struct *tsk)
 584 {
 585         int i, call_panic = 0;
 586         struct audit_buffer *ab;
 587         struct audit_aux_data *aux;
 588         const char *tty;
 589
 590         /* tsk == current */
 591
 592         ab = audit_log_start(context, GFP_KERNEL, AUDIT_SYSCALL);
 593         if (!ab)
 594                 return;         /* audit_panic has been called */
 595         audit_log_format(ab, "arch=%x syscall=%d",
 596                          context->arch, context->major);
 597         if (context->personality != PER_LINUX)
 598                 audit_log_format(ab, " per=%lx", context->personality);
 599         if (context->return_valid)
 600                 audit_log_format(ab, " success=%s exit=%ld",
 601                                  (context->return_valid==AUDITSC_SUCCESS)?"yes":"no",
 602                                  context->return_code);
 603         if (tsk->signal && tsk->signal->tty && tsk->signal->tty->name)
 604                 tty = tsk->signal->tty->name;
 605         else
 606                 tty = "(none)";
 607         audit_log_format(ab,
 608                   " a0=%lx a1=%lx a2=%lx a3=%lx items=%d"
 609                   " pid=%d auid=%u uid=%u gid=%u"
 610                   " euid=%u suid=%u fsuid=%u"
 611                   " egid=%u sgid=%u fsgid=%u tty=%s",
 612                   context->argv[0],
 613                   context->argv[1],
 614                   context->argv[2],
 615                   context->argv[3],
 616                   context->name_count,
 617                   context->pid,
 618                   context->loginuid,
 619                   context->uid,
 620                   context->gid,
 621                   context->euid, context->suid, context->fsuid,
 622                   context->egid, context->sgid, context->fsgid, tty);
 623         audit_log_task_info(ab, tsk);
 624         audit_log_end(ab);
 625
 626         for (aux = context->aux; aux; aux = aux->next) {
 627
 628                 ab = audit_log_start(context, GFP_KERNEL, aux->type);
 629                 if (!ab)
 630                         continue; /* audit_panic has been called */
 631
 632                 switch (aux->type) {
 633                 case AUDIT_IPC: {
 634                         struct audit_aux_data_ipcctl *axi = (void *)aux;
 635                         audit_log_format(ab,
 636                                  " qbytes=%lx iuid=%u igid=%u mode=%x",
 637                                  axi->qbytes, axi->uid, axi->gid, axi->mode);
 638                         if (axi->osid != 0) {
 639                                 char *ctx = NULL;
 640                                 u32 len;
 641                                 if (selinux_ctxid_to_string(
 642                                                 axi->osid, &ctx, &len)) {
 643                                         audit_log_format(ab, " osid=%u",
 644                                                         axi->osid);
 645                                         call_panic = 1;
 646                                 } else
 647                                         audit_log_format(ab, " obj=%s", ctx);
 648                                 kfree(ctx);
 649                         }
 650                         break; }
 651
 652                 case AUDIT_IPC_SET_PERM: {
 653                         struct audit_aux_data_ipcctl *axi = (void *)aux;
 654                         audit_log_format(ab,
 655                                 " new qbytes=%lx new iuid=%u new igid=%u new mode=%x",
 656                                 axi->qbytes, axi->uid, axi->gid, axi->mode);
 657                         if (axi->osid != 0) {
 658                                 char *ctx = NULL;
 659                                 u32 len;
 660                                 if (selinux_ctxid_to_string(
 661                                                 axi->osid, &ctx, &len)) {
 662                                         audit_log_format(ab, " osid=%u",
 663                                                         axi->osid);
 664                                         call_panic = 1;
 665                                 } else
 666                                         audit_log_format(ab, " obj=%s", ctx);
 667                                 kfree(ctx);
 668                         }
 669                         break; }
 670
 671                 case AUDIT_SOCKETCALL: {
 672                         int i;
 673                         struct audit_aux_data_socketcall *axs = (void *)aux;
 674                         audit_log_format(ab, "nargs=%d", axs->nargs);
 675                         for (i=0; i<axs->nargs; i++)
 676                                 audit_log_format(ab, " a%d=%lx", i, axs->args[i]);
 677                         break; }
 678
 679                 case AUDIT_SOCKADDR: {
 680                         struct audit_aux_data_sockaddr *axs = (void *)aux;
 681
 682                         audit_log_format(ab, "saddr=");
 683                         audit_log_hex(ab, axs->a, axs->len);
 684                         break; }
 685
 686                 case AUDIT_AVC_PATH: {
 687                         struct audit_aux_data_path *axi = (void *)aux;
 688                         audit_log_d_path(ab, "path=", axi->dentry, axi->mnt);
 689                         break; }
 690
 691                 }
 692                 audit_log_end(ab);
 693         }
 694
 695         if (context->pwd && context->pwdmnt) {
 696                 ab = audit_log_start(context, GFP_KERNEL, AUDIT_CWD);
 697                 if (ab) {
 698                         audit_log_d_path(ab, "cwd=", context->pwd, context->pwdmnt);
 699                         audit_log_end(ab);
 700                 }
 701         }
 702         for (i = 0; i < context->name_count; i++) {
 703                 unsigned long ino  = context->names[i].ino;
 704                 unsigned long pino = context->names[i].pino;
 705
 706                 ab = audit_log_start(context, GFP_KERNEL, AUDIT_PATH);
 707                 if (!ab)
 708                         continue; /* audit_panic has been called */
 709
 710                 audit_log_format(ab, "item=%d", i);
 711
 712                 audit_log_format(ab, " name=");
 713                 if (context->names[i].name)
 714                         audit_log_untrustedstring(ab, context->names[i].name);
 715                 else
 716                         audit_log_format(ab, "(null)");
 717
 718                 if (pino != (unsigned long)-1)
 719                         audit_log_format(ab, " parent=%lu",  pino);
 720                 if (ino != (unsigned long)-1)
 721                         audit_log_format(ab, " inode=%lu",  ino);
 722                 if ((pino != (unsigned long)-1) || (ino != (unsigned long)-1))
 723                         audit_log_format(ab, " dev=%02x:%02x mode=%#o"
 724                                          " ouid=%u ogid=%u rdev=%02x:%02x",
 725                                          MAJOR(context->names[i].dev),
 726                                          MINOR(context->names[i].dev),
 727                                          context->names[i].mode,
 728                                          context->names[i].uid,
 729                                          context->names[i].gid,
 730                                          MAJOR(context->names[i].rdev),
 731                                          MINOR(context->names[i].rdev));
 732                 if (context->names[i].osid != 0) {
 733                         char *ctx = NULL;
 734                         u32 len;
 735                         if (selinux_ctxid_to_string(
 736                                 context->names[i].osid, &ctx, &len)) {
 737                                 audit_log_format(ab, " osid=%u",
 738                                                 context->names[i].osid);
 739                                 call_panic = 2;
 740                         } else
 741                                 audit_log_format(ab, " obj=%s", ctx);
 742                         kfree(ctx);
 743                 }
 744
 745                 audit_log_end(ab);
 746         }
 747         if (call_panic)
 748                 audit_panic("error converting sid to string");
 749 }
 750
 751 /**
 752  * audit_free - free a per-task audit context
 753  * @tsk: task whose audit context block to free
 754  *
 755  * Called from copy_process and do_exit
 756  */
 757 void audit_free(struct task_struct *tsk)
 758 {
 759         struct audit_context *context;
 760
 761         context = audit_get_context(tsk, 0, 0);
 762         if (likely(!context))
 763                 return;
 764
 765         /* Check for system calls that do not go through the exit
 766          * function (e.g., exit_group), then free context block.
 767          * We use GFP_ATOMIC here because we might be doing this
 768          * in the context of the idle thread */
 769         /* that can happen only if we are called from do_exit() */
 770         if (context->in_syscall && context->auditable)
 771                 audit_log_exit(context, tsk);
 772
 773         audit_free_context(context);
 774 }
 775
 776 /**
 777  * audit_syscall_entry - fill in an audit record at syscall entry
 778  * @tsk: task being audited
 779  * @arch: architecture type
 780  * @major: major syscall type (function)
 781  * @a1: additional syscall register 1
 782  * @a2: additional syscall register 2
 783  * @a3: additional syscall register 3
 784  * @a4: additional syscall register 4
 785  *
 786  * Fill in audit context at syscall entry.  This only happens if the
 787  * audit context was created when the task was created and the state or
 788  * filters demand the audit context be built.  If the state from the
 789  * per-task filter or from the per-syscall filter is AUDIT_RECORD_CONTEXT,
 790  * then the record will be written at syscall exit time (otherwise, it
 791  * will only be written if another part of the kernel requests that it
 792  * be written).
 793  */
 794 void audit_syscall_entry(int arch, int major,
 795                          unsigned long a1, unsigned long a2,
 796                          unsigned long a3, unsigned long a4)
 797 {
 798         struct task_struct *tsk = current;
 799         struct audit_context *context = tsk->audit_context;
 800         enum audit_state     state;
 801
 802         BUG_ON(!context);
 803
 804         /*
 805          * This happens only on certain architectures that make system
 806          * calls in kernel_thread via the entry.S interface, instead of
 807          * with direct calls.  (If you are porting to a new
 808          * architecture, hitting this condition can indicate that you
 809          * got the _exit/_leave calls backward in entry.S.)
 810          *
 811          * i386     no
 812          * x86_64   no
 813          * ppc64    yes (see arch/powerpc/platforms/iseries/misc.S)
 814          *
 815          * This also happens with vm86 emulation in a non-nested manner
 816          * (entries without exits), so this case must be caught.
 817          */
 818         if (context->in_syscall) {
 819                 struct audit_context *newctx;
 820
 821 #if AUDIT_DEBUG
 822                 printk(KERN_ERR
 823                        "audit(:%d) pid=%d in syscall=%d;"
 824                        " entering syscall=%d\n",
 825                        context->serial, tsk->pid, context->major, major);
 826 #endif
 827                 newctx = audit_alloc_context(context->state);
 828                 if (newctx) {
 829                         newctx->previous   = context;
 830                         context            = newctx;
 831                         tsk->audit_context = newctx;
 832                 } else  {
 833                         /* If we can't alloc a new context, the best we
 834                          * can do is to leak memory (any pending putname
 835                          * will be lost).  The only other alternative is
 836                          * to abandon auditing. */
 837                         audit_zero_context(context, context->state);
 838                 }
 839         }
 840         BUG_ON(context->in_syscall || context->name_count);
 841
 842         if (!audit_enabled)
 843                 return;
 844
 845         context->arch       = arch;
 846         context->major      = major;
 847         context->argv[0]    = a1;
 848         context->argv[1]    = a2;
 849         context->argv[2]    = a3;
 850         context->argv[3]    = a4;
 851
 852         state = context->state;
 853         if (state == AUDIT_SETUP_CONTEXT || state == AUDIT_BUILD_CONTEXT)
 854                 state = audit_filter_syscall(tsk, context, &audit_filter_list[AUDIT_FILTER_ENTRY]);
 855         if (likely(state == AUDIT_DISABLED))
 856                 return;
 857
 858         context->serial     = 0;
 859         context->ctime      = CURRENT_TIME;
 860         context->in_syscall = 1;
 861         context->auditable  = !!(state == AUDIT_RECORD_CONTEXT);
 862 }
 863
 864 /**
 865  * audit_syscall_exit - deallocate audit context after a system call
 866  * @tsk: task being audited
 867  * @valid: success/failure flag
 868  * @return_code: syscall return value
 869  *
 870  * Tear down after system call.  If the audit context has been marked as
 871  * auditable (either because of the AUDIT_RECORD_CONTEXT state from
 872  * filtering, or because some other part of the kernel write an audit
 873  * message), then write out the syscall information.  In call cases,
 874  * free the names stored from getname().
 875  */
 876 void audit_syscall_exit(int valid, long return_code)
 877 {
 878         struct task_struct *tsk = current;
 879         struct audit_context *context;
 880
 881         context = audit_get_context(tsk, valid, return_code);
 882
 883         if (likely(!context))
 884                 return;
 885
 886         if (context->in_syscall && context->auditable)
 887                 audit_log_exit(context, tsk);
 888
 889         context->in_syscall = 0;
 890         context->auditable  = 0;
 891
 892         if (context->previous) {
 893                 struct audit_context *new_context = context->previous;
 894                 context->previous  = NULL;
 895                 audit_free_context(context);
 896                 tsk->audit_context = new_context;
 897         } else {
 898                 audit_free_names(context);
 899                 audit_free_aux(context);
 900                 tsk->audit_context = context;
 901         }
 902 }
 903
 904 /**
 905  * audit_getname - add a name to the list
 906  * @name: name to add
 907  *
 908  * Add a name to the list of audit names for this context.
 909  * Called from fs/namei.c:getname().
 910  */
 911 void audit_getname(const char *name)
 912 {
 913         struct audit_context *context = current->audit_context;
 914
 915         if (!context || IS_ERR(name) || !name)
 916                 return;
 917
 918         if (!context->in_syscall) {
 919 #if AUDIT_DEBUG == 2
 920                 printk(KERN_ERR "%s:%d(:%d): ignoring getname(%p)\n",
 921                        __FILE__, __LINE__, context->serial, name);
 922                 dump_stack();
 923 #endif
 924                 return;
 925         }
 926         BUG_ON(context->name_count >= AUDIT_NAMES);
 927         context->names[context->name_count].name = name;
 928         context->names[context->name_count].ino  = (unsigned long)-1;
 929         ++context->name_count;
 930         if (!context->pwd) {
 931                 read_lock(&current->fs->lock);
 932                 context->pwd = dget(current->fs->pwd);
 933                 context->pwdmnt = mntget(current->fs->pwdmnt);
 934                 read_unlock(&current->fs->lock);
 935         }
 936
 937 }
 938
 939 /* audit_putname - intercept a putname request
 940  * @name: name to intercept and delay for putname
 941  *
 942  * If we have stored the name from getname in the audit context,
 943  * then we delay the putname until syscall exit.
 944  * Called from include/linux/fs.h:putname().
 945  */
 946 void audit_putname(const char *name)
 947 {
 948         struct audit_context *context = current->audit_context;
 949
 950         BUG_ON(!context);
 951         if (!context->in_syscall) {
 952 #if AUDIT_DEBUG == 2
 953                 printk(KERN_ERR "%s:%d(:%d): __putname(%p)\n",
 954                        __FILE__, __LINE__, context->serial, name);
 955                 if (context->name_count) {
 956                         int i;
 957                         for (i = 0; i < context->name_count; i++)
 958                                 printk(KERN_ERR "name[%d] = %p = %s\n", i,
 959                                        context->names[i].name,
 960                                        context->names[i].name ?: "(null)");
 961                 }
 962 #endif
 963                 __putname(name);
 964         }
 965 #if AUDIT_DEBUG
 966         else {
 967                 ++context->put_count;
 968                 if (context->put_count > context->name_count) {
 969                         printk(KERN_ERR "%s:%d(:%d): major=%d"
 970                                " in_syscall=%d putname(%p) name_count=%d"
 971                                " put_count=%d\n",
 972                                __FILE__, __LINE__,
 973                                context->serial, context->major,
 974                                context->in_syscall, name, context->name_count,
 975                                context->put_count);
 976                         dump_stack();
 977                 }
 978         }
 979 #endif
 980 }
 981
 982 static void audit_inode_context(int idx, const struct inode *inode)
 983 {
 984         struct audit_context *context = current->audit_context;
 985
 986         selinux_get_inode_sid(inode, &context->names[idx].osid);
 987 }
 988
 989
 990 /**
 991  * audit_inode - store the inode and device from a lookup
 992  * @name: name being audited
 993  * @inode: inode being audited
 994  * @flags: lookup flags (as used in path_lookup())
 995  *
 996  * Called from fs/namei.c:path_lookup().
 997  */
 998 void __audit_inode(const char *name, const struct inode *inode, unsigned flags)
 999 {
1000         int idx;
1001         struct audit_context *context = current->audit_context;
1002
1003         if (!context->in_syscall)
1004                 return;
1005         if (context->name_count
1006             && context->names[context->name_count-1].name
1007             && context->names[context->name_count-1].name == name)
1008                 idx = context->name_count - 1;
1009         else if (context->name_count > 1
1010                  && context->names[context->name_count-2].name
1011                  && context->names[context->name_count-2].name == name)
1012                 idx = context->name_count - 2;
1013         else {
1014                 /* FIXME: how much do we care about inodes that have no
1015                  * associated name? */
1016                 if (context->name_count >= AUDIT_NAMES - AUDIT_NAMES_RESERVED)
1017                         return;
1018                 idx = context->name_count++;
1019                 context->names[idx].name = NULL;
1020 #if AUDIT_DEBUG
1021                 ++context->ino_count;
1022 #endif
1023         }
1024         context->names[idx].dev   = inode->i_sb->s_dev;
1025         context->names[idx].mode  = inode->i_mode;
1026         context->names[idx].uid   = inode->i_uid;
1027         context->names[idx].gid   = inode->i_gid;
1028         context->names[idx].rdev  = inode->i_rdev;
1029         audit_inode_context(idx, inode);
1030         if ((flags & LOOKUP_PARENT) && (strcmp(name, "/") != 0) &&
1031             (strcmp(name, ".") != 0)) {
1032                 context->names[idx].ino   = (unsigned long)-1;
1033                 context->names[idx].pino  = inode->i_ino;
1034         } else {
1035                 context->names[idx].ino   = inode->i_ino;
1036                 context->names[idx].pino  = (unsigned long)-1;
1037         }
1038 }
1039
1040 /**
1041  * audit_inode_child - collect inode info for created/removed objects
1042  * @dname: inode's dentry name
1043  * @inode: inode being audited
1044  * @pino: inode number of dentry parent
1045  *
1046  * For syscalls that create or remove filesystem objects, audit_inode
1047  * can only collect information for the filesystem object's parent.
1048  * This call updates the audit context with the child's information.
1049  * Syscalls that create a new filesystem object must be hooked after
1050  * the object is created.  Syscalls that remove a filesystem object
1051  * must be hooked prior, in order to capture the target inode during
1052  * unsuccessful attempts.
1053  */
1054 void __audit_inode_child(const char *dname, const struct inode *inode,
1055                          unsigned long pino)
1056 {
1057         int idx;
1058         struct audit_context *context = current->audit_context;
1059
1060         if (!context->in_syscall)
1061                 return;
1062
1063         /* determine matching parent */
1064         if (dname)
1065                 for (idx = 0; idx < context->name_count; idx++)
1066                         if (context->names[idx].pino == pino) {
1067                                 const char *n;
1068                                 const char *name = context->names[idx].name;
1069                                 int dlen = strlen(dname);
1070                                 int nlen = name ? strlen(name) : 0;
1071
1072                                 if (nlen < dlen)
1073                                         continue;
1074
1075                                 /* disregard trailing slashes */
1076                                 n = name + nlen - 1;
1077                                 while ((*n == '/') && (n > name))
1078                                         n--;
1079
1080                                 /* find last path component */
1081                                 n = n - dlen + 1;
1082                                 if (n < name)
1083                                         continue;
1084                                 else if (n > name) {
1085                                         if (*--n != '/')
1086                                                 continue;
1087                                         else
1088                                                 n++;
1089                                 }
1090
1091                                 if (strncmp(n, dname, dlen) == 0)
1092                                         goto update_context;
1093                         }
1094
1095         /* catch-all in case match not found */
1096         idx = context->name_count++;
1097         context->names[idx].name  = NULL;
1098         context->names[idx].pino  = pino;
1099 #if AUDIT_DEBUG
1100         context->ino_count++;
1101 #endif
1102
1103 update_context:
1104         if (inode) {
1105                 context->names[idx].ino   = inode->i_ino;
1106                 context->names[idx].dev   = inode->i_sb->s_dev;
1107                 context->names[idx].mode  = inode->i_mode;
1108                 context->names[idx].uid   = inode->i_uid;
1109                 context->names[idx].gid   = inode->i_gid;
1110                 context->names[idx].rdev  = inode->i_rdev;
1111                 audit_inode_context(idx, inode);
1112         }
1113 }
1114
1115 /**
1116  * auditsc_get_stamp - get local copies of audit_context values
1117  * @ctx: audit_context for the task
1118  * @t: timespec to store time recorded in the audit_context
1119  * @serial: serial value that is recorded in the audit_context
1120  *
1121  * Also sets the context as auditable.
1122  */
1123 void auditsc_get_stamp(struct audit_context *ctx,
1124                        struct timespec *t, unsigned int *serial)
1125 {
1126         if (!ctx->serial)
1127                 ctx->serial = audit_serial();
1128         t->tv_sec  = ctx->ctime.tv_sec;
1129         t->tv_nsec = ctx->ctime.tv_nsec;
1130         *serial    = ctx->serial;
1131         ctx->auditable = 1;
1132 }
1133
1134 /**
1135  * audit_set_loginuid - set a task's audit_context loginuid
1136  * @task: task whose audit context is being modified
1137  * @loginuid: loginuid value
1138  *
1139  * Returns 0.
1140  *
1141  * Called (set) from fs/proc/base.c::proc_loginuid_write().
1142  */
1143 int audit_set_loginuid(struct task_struct *task, uid_t loginuid)
1144 {
1145         if (task->audit_context) {
1146                 struct audit_buffer *ab;
1147
1148                 ab = audit_log_start(NULL, GFP_KERNEL, AUDIT_LOGIN);
1149                 if (ab) {
1150                         audit_log_format(ab, "login pid=%d uid=%u "
1151                                 "old auid=%u new auid=%u",
1152                                 task->pid, task->uid,
1153                                 task->audit_context->loginuid, loginuid);
1154                         audit_log_end(ab);
1155                 }
1156                 task->audit_context->loginuid = loginuid;
1157         }
1158         return 0;
1159 }
1160
1161 /**
1162  * audit_get_loginuid - get the loginuid for an audit_context
1163  * @ctx: the audit_context
1164  *
1165  * Returns the context's loginuid or -1 if @ctx is NULL.
1166  */
1167 uid_t audit_get_loginuid(struct audit_context *ctx)
1168 {
1169         return ctx ? ctx->loginuid : -1;
1170 }
1171
1172 /**
1173  * audit_ipc_obj - record audit data for ipc object
1174  * @ipcp: ipc permissions
1175  *
1176  * Returns 0 for success or NULL context or < 0 on error.
1177  */
1178 int audit_ipc_obj(struct kern_ipc_perm *ipcp)
1179 {
1180         struct audit_aux_data_ipcctl *ax;
1181         struct audit_context *context = current->audit_context;
1182
1183         if (likely(!context))
1184                 return 0;
1185
1186         ax = kmalloc(sizeof(*ax), GFP_ATOMIC);
1187         if (!ax)
1188                 return -ENOMEM;
1189
1190         ax->uid = ipcp->uid;
1191         ax->gid = ipcp->gid;
1192         ax->mode = ipcp->mode;
1193         selinux_get_ipc_sid(ipcp, &ax->osid);
1194
1195         ax->d.type = AUDIT_IPC;
1196         ax->d.next = context->aux;
1197         context->aux = (void *)ax;
1198         return 0;
1199 }
1200
1201 /**
1202  * audit_ipc_set_perm - record audit data for new ipc permissions
1203  * @qbytes: msgq bytes
1204  * @uid: msgq user id
1205  * @gid: msgq group id
1206  * @mode: msgq mode (permissions)
1207  *
1208  * Returns 0 for success or NULL context or < 0 on error.
1209  */
1210 int audit_ipc_set_perm(unsigned long qbytes, uid_t uid, gid_t gid, mode_t mode, struct kern_ipc_perm *ipcp)
1211 {
1212         struct audit_aux_data_ipcctl *ax;
1213         struct audit_context *context = current->audit_context;
1214
1215         if (likely(!context))
1216                 return 0;
1217
1218         ax = kmalloc(sizeof(*ax), GFP_ATOMIC);
1219         if (!ax)
1220                 return -ENOMEM;
1221
1222         ax->qbytes = qbytes;
1223         ax->uid = uid;
1224         ax->gid = gid;
1225         ax->mode = mode;
1226         selinux_get_ipc_sid(ipcp, &ax->osid);
1227
1228         ax->d.type = AUDIT_IPC_SET_PERM;
1229         ax->d.next = context->aux;
1230         context->aux = (void *)ax;
1231         return 0;
1232 }
1233
1234 /**
1235  * audit_socketcall - record audit data for sys_socketcall
1236  * @nargs: number of args
1237  * @args: args array
1238  *
1239  * Returns 0 for success or NULL context or < 0 on error.
1240  */
1241 int audit_socketcall(int nargs, unsigned long *args)
1242 {
1243         struct audit_aux_data_socketcall *ax;
1244         struct audit_context *context = current->audit_context;
1245
1246         if (likely(!context))
1247                 return 0;
1248
1249         ax = kmalloc(sizeof(*ax) + nargs * sizeof(unsigned long), GFP_KERNEL);
1250         if (!ax)
1251                 return -ENOMEM;
1252
1253         ax->nargs = nargs;
1254         memcpy(ax->args, args, nargs * sizeof(unsigned long));
1255
1256         ax->d.type = AUDIT_SOCKETCALL;
1257         ax->d.next = context->aux;
1258         context->aux = (void *)ax;
1259         return 0;
1260 }
1261
1262 /**
1263  * audit_sockaddr - record audit data for sys_bind, sys_connect, sys_sendto
1264  * @len: data length in user space
1265  * @a: data address in kernel space
1266  *
1267  * Returns 0 for success or NULL context or < 0 on error.
1268  */
1269 int audit_sockaddr(int len, void *a)
1270 {
1271         struct audit_aux_data_sockaddr *ax;
1272         struct audit_context *context = current->audit_context;
1273
1274         if (likely(!context))
1275                 return 0;
1276
1277         ax = kmalloc(sizeof(*ax) + len, GFP_KERNEL);
1278         if (!ax)
1279                 return -ENOMEM;
1280
1281         ax->len = len;
1282         memcpy(ax->a, a, len);
1283
1284         ax->d.type = AUDIT_SOCKADDR;
1285         ax->d.next = context->aux;
1286         context->aux = (void *)ax;
1287         return 0;
1288 }
1289
1290 /**
1291  * audit_avc_path - record the granting or denial of permissions
1292  * @dentry: dentry to record
1293  * @mnt: mnt to record
1294  *
1295  * Returns 0 for success or NULL context or < 0 on error.
1296  *
1297  * Called from security/selinux/avc.c::avc_audit()
1298  */
1299 int audit_avc_path(struct dentry *dentry, struct vfsmount *mnt)
1300 {
1301         struct audit_aux_data_path *ax;
1302         struct audit_context *context = current->audit_context;
1303
1304         if (likely(!context))
1305                 return 0;
1306
1307         ax = kmalloc(sizeof(*ax), GFP_ATOMIC);
1308         if (!ax)
1309                 return -ENOMEM;
1310
1311         ax->dentry = dget(dentry);
1312         ax->mnt = mntget(mnt);
1313
1314         ax->d.type = AUDIT_AVC_PATH;
1315         ax->d.next = context->aux;
1316         context->aux = (void *)ax;
1317         return 0;
1318 }
1319
1320 /**
1321  * audit_signal_info - record signal info for shutting down audit subsystem
1322  * @sig: signal value
1323  * @t: task being signaled
1324  *
1325  * If the audit subsystem is being terminated, record the task (pid)
1326  * and uid that is doing that.
1327  */
1328 void audit_signal_info(int sig, struct task_struct *t)
1329 {
1330         extern pid_t audit_sig_pid;
1331         extern uid_t audit_sig_uid;
1332
1333         if (unlikely(audit_pid && t->tgid == audit_pid)) {
1334                 if (sig == SIGTERM || sig == SIGHUP) {
1335                         struct audit_context *ctx = current->audit_context;
1336                         audit_sig_pid = current->pid;
1337                         if (ctx)
1338                                 audit_sig_uid = ctx->loginuid;
1339                         else
1340                                 audit_sig_uid = current->uid;
1341                 }
1342         }
1343 }