linux/kernel/auditsc.c
<<
>>
Prefs
   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, 2006 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 * POSIX message queue support added by George Wilson <ltcgcw@us.ibm.com>,
  33 * 2006.
  34 *
  35 * The support of additional filter rules compares (>, <, >=, <=) was
  36 * added by Dustin Kirkland <dustin.kirkland@us.ibm.com>, 2005.
  37 *
  38 * Modified by Amy Griffis <amy.griffis@hp.com> to collect additional
  39 * filesystem information.
  40 *
  41 * Subject and object context labeling support added by <danjones@us.ibm.com>
  42 * and <dustin.kirkland@us.ibm.com> for LSPP certification compliance.
  43 */
  44
  45#include <linux/init.h>
  46#include <asm/types.h>
  47#include <linux/atomic.h>
  48#include <linux/fs.h>
  49#include <linux/namei.h>
  50#include <linux/mm.h>
  51#include <linux/module.h>
  52#include <linux/slab.h>
  53#include <linux/mount.h>
  54#include <linux/socket.h>
  55#include <linux/mqueue.h>
  56#include <linux/audit.h>
  57#include <linux/personality.h>
  58#include <linux/time.h>
  59#include <linux/netlink.h>
  60#include <linux/compiler.h>
  61#include <asm/unistd.h>
  62#include <linux/security.h>
  63#include <linux/list.h>
  64#include <linux/tty.h>
  65#include <linux/binfmts.h>
  66#include <linux/highmem.h>
  67#include <linux/syscalls.h>
  68#include <linux/capability.h>
  69#include <linux/fs_struct.h>
  70
  71#include "audit.h"
  72
  73/* AUDIT_NAMES is the number of slots we reserve in the audit_context
  74 * for saving names from getname(). */
  75#define AUDIT_NAMES    20
  76
  77/* Indicates that audit should log the full pathname. */
  78#define AUDIT_NAME_FULL -1
  79
  80/* no execve audit message should be longer than this (userspace limits) */
  81#define MAX_EXECVE_AUDIT_LEN 7500
  82
  83/* number of audit rules */
  84int audit_n_rules;
  85
  86/* determines whether we collect data for signals sent */
  87int audit_signals;
  88
  89struct audit_cap_data {
  90        kernel_cap_t            permitted;
  91        kernel_cap_t            inheritable;
  92        union {
  93                unsigned int    fE;             /* effective bit of a file capability */
  94                kernel_cap_t    effective;      /* effective set of a process */
  95        };
  96};
  97
  98/* When fs/namei.c:getname() is called, we store the pointer in name and
  99 * we don't let putname() free it (instead we free all of the saved
 100 * pointers at syscall exit time).
 101 *
 102 * Further, in fs/namei.c:path_lookup() we store the inode and device. */
 103struct audit_names {
 104        const char      *name;
 105        int             name_len;       /* number of name's characters to log */
 106        unsigned        name_put;       /* call __putname() for this name */
 107        unsigned long   ino;
 108        dev_t           dev;
 109        umode_t         mode;
 110        uid_t           uid;
 111        gid_t           gid;
 112        dev_t           rdev;
 113        u32             osid;
 114        struct audit_cap_data fcap;
 115        unsigned int    fcap_ver;
 116};
 117
 118struct audit_aux_data {
 119        struct audit_aux_data   *next;
 120        int                     type;
 121};
 122
 123#define AUDIT_AUX_IPCPERM       0
 124
 125/* Number of target pids per aux struct. */
 126#define AUDIT_AUX_PIDS  16
 127
 128struct audit_aux_data_execve {
 129        struct audit_aux_data   d;
 130        int argc;
 131        int envc;
 132        struct mm_struct *mm;
 133};
 134
 135struct audit_aux_data_pids {
 136        struct audit_aux_data   d;
 137        pid_t                   target_pid[AUDIT_AUX_PIDS];
 138        uid_t                   target_auid[AUDIT_AUX_PIDS];
 139        uid_t                   target_uid[AUDIT_AUX_PIDS];
 140        unsigned int            target_sessionid[AUDIT_AUX_PIDS];
 141        u32                     target_sid[AUDIT_AUX_PIDS];
 142        char                    target_comm[AUDIT_AUX_PIDS][TASK_COMM_LEN];
 143        int                     pid_count;
 144};
 145
 146struct audit_aux_data_bprm_fcaps {
 147        struct audit_aux_data   d;
 148        struct audit_cap_data   fcap;
 149        unsigned int            fcap_ver;
 150        struct audit_cap_data   old_pcap;
 151        struct audit_cap_data   new_pcap;
 152};
 153
 154struct audit_aux_data_capset {
 155        struct audit_aux_data   d;
 156        pid_t                   pid;
 157        struct audit_cap_data   cap;
 158};
 159
 160struct audit_tree_refs {
 161        struct audit_tree_refs *next;
 162        struct audit_chunk *c[31];
 163};
 164
 165/* The per-task audit context. */
 166struct audit_context {
 167        int                 dummy;      /* must be the first element */
 168        int                 in_syscall; /* 1 if task is in a syscall */
 169        enum audit_state    state, current_state;
 170        unsigned int        serial;     /* serial number for record */
 171        int                 major;      /* syscall number */
 172        struct timespec     ctime;      /* time of syscall entry */
 173        unsigned long       argv[4];    /* syscall arguments */
 174        long                return_code;/* syscall return code */
 175        u64                 prio;
 176        int                 return_valid; /* return code is valid */
 177        int                 name_count;
 178        struct audit_names  names[AUDIT_NAMES];
 179        char *              filterkey;  /* key for rule that triggered record */
 180        struct path         pwd;
 181        struct audit_context *previous; /* For nested syscalls */
 182        struct audit_aux_data *aux;
 183        struct audit_aux_data *aux_pids;
 184        struct sockaddr_storage *sockaddr;
 185        size_t sockaddr_len;
 186                                /* Save things to print about task_struct */
 187        pid_t               pid, ppid;
 188        uid_t               uid, euid, suid, fsuid;
 189        gid_t               gid, egid, sgid, fsgid;
 190        unsigned long       personality;
 191        int                 arch;
 192
 193        pid_t               target_pid;
 194        uid_t               target_auid;
 195        uid_t               target_uid;
 196        unsigned int        target_sessionid;
 197        u32                 target_sid;
 198        char                target_comm[TASK_COMM_LEN];
 199
 200        struct audit_tree_refs *trees, *first_trees;
 201        struct list_head killed_trees;
 202        int tree_count;
 203
 204        int type;
 205        union {
 206                struct {
 207                        int nargs;
 208                        long args[6];
 209                } socketcall;
 210                struct {
 211                        uid_t                   uid;
 212                        gid_t                   gid;
 213                        mode_t                  mode;
 214                        u32                     osid;
 215                        int                     has_perm;
 216                        uid_t                   perm_uid;
 217                        gid_t                   perm_gid;
 218                        mode_t                  perm_mode;
 219                        unsigned long           qbytes;
 220                } ipc;
 221                struct {
 222                        mqd_t                   mqdes;
 223                        struct mq_attr          mqstat;
 224                } mq_getsetattr;
 225                struct {
 226                        mqd_t                   mqdes;
 227                        int                     sigev_signo;
 228                } mq_notify;
 229                struct {
 230                        mqd_t                   mqdes;
 231                        size_t                  msg_len;
 232                        unsigned int            msg_prio;
 233                        struct timespec         abs_timeout;
 234                } mq_sendrecv;
 235                struct {
 236                        int                     oflag;
 237                        mode_t                  mode;
 238                        struct mq_attr          attr;
 239                } mq_open;
 240                struct {
 241                        pid_t                   pid;
 242                        struct audit_cap_data   cap;
 243                } capset;
 244                struct {
 245                        int                     fd;
 246                        int                     flags;
 247                } mmap;
 248        };
 249        int fds[2];
 250
 251#if AUDIT_DEBUG
 252        int                 put_count;
 253        int                 ino_count;
 254#endif
 255};
 256
 257static inline int open_arg(int flags, int mask)
 258{
 259        int n = ACC_MODE(flags);
 260        if (flags & (O_TRUNC | O_CREAT))
 261                n |= AUDIT_PERM_WRITE;
 262        return n & mask;
 263}
 264
 265static int audit_match_perm(struct audit_context *ctx, int mask)
 266{
 267        unsigned n;
 268        if (unlikely(!ctx))
 269                return 0;
 270        n = ctx->major;
 271
 272        switch (audit_classify_syscall(ctx->arch, n)) {
 273        case 0: /* native */
 274                if ((mask & AUDIT_PERM_WRITE) &&
 275                     audit_match_class(AUDIT_CLASS_WRITE, n))
 276                        return 1;
 277                if ((mask & AUDIT_PERM_READ) &&
 278                     audit_match_class(AUDIT_CLASS_READ, n))
 279                        return 1;
 280                if ((mask & AUDIT_PERM_ATTR) &&
 281                     audit_match_class(AUDIT_CLASS_CHATTR, n))
 282                        return 1;
 283                return 0;
 284        case 1: /* 32bit on biarch */
 285                if ((mask & AUDIT_PERM_WRITE) &&
 286                     audit_match_class(AUDIT_CLASS_WRITE_32, n))
 287                        return 1;
 288                if ((mask & AUDIT_PERM_READ) &&
 289                     audit_match_class(AUDIT_CLASS_READ_32, n))
 290                        return 1;
 291                if ((mask & AUDIT_PERM_ATTR) &&
 292                     audit_match_class(AUDIT_CLASS_CHATTR_32, n))
 293                        return 1;
 294                return 0;
 295        case 2: /* open */
 296                return mask & ACC_MODE(ctx->argv[1]);
 297        case 3: /* openat */
 298                return mask & ACC_MODE(ctx->argv[2]);
 299        case 4: /* socketcall */
 300                return ((mask & AUDIT_PERM_WRITE) && ctx->argv[0] == SYS_BIND);
 301        case 5: /* execve */
 302                return mask & AUDIT_PERM_EXEC;
 303        default:
 304                return 0;
 305        }
 306}
 307
 308static int audit_match_filetype(struct audit_context *ctx, int which)
 309{
 310        unsigned index = which & ~S_IFMT;
 311        mode_t mode = which & S_IFMT;
 312
 313        if (unlikely(!ctx))
 314                return 0;
 315
 316        if (index >= ctx->name_count)
 317                return 0;
 318        if (ctx->names[index].ino == -1)
 319                return 0;
 320        if ((ctx->names[index].mode ^ mode) & S_IFMT)
 321                return 0;
 322        return 1;
 323}
 324
 325/*
 326 * We keep a linked list of fixed-sized (31 pointer) arrays of audit_chunk *;
 327 * ->first_trees points to its beginning, ->trees - to the current end of data.
 328 * ->tree_count is the number of free entries in array pointed to by ->trees.
 329 * Original condition is (NULL, NULL, 0); as soon as it grows we never revert to NULL,
 330 * "empty" becomes (p, p, 31) afterwards.  We don't shrink the list (and seriously,
 331 * it's going to remain 1-element for almost any setup) until we free context itself.
 332 * References in it _are_ dropped - at the same time we free/drop aux stuff.
 333 */
 334
 335#ifdef CONFIG_AUDIT_TREE
 336static void audit_set_auditable(struct audit_context *ctx)
 337{
 338        if (!ctx->prio) {
 339                ctx->prio = 1;
 340                ctx->current_state = AUDIT_RECORD_CONTEXT;
 341        }
 342}
 343
 344static int put_tree_ref(struct audit_context *ctx, struct audit_chunk *chunk)
 345{
 346        struct audit_tree_refs *p = ctx->trees;
 347        int left = ctx->tree_count;
 348        if (likely(left)) {
 349                p->c[--left] = chunk;
 350                ctx->tree_count = left;
 351                return 1;
 352        }
 353        if (!p)
 354                return 0;
 355        p = p->next;
 356        if (p) {
 357                p->c[30] = chunk;
 358                ctx->trees = p;
 359                ctx->tree_count = 30;
 360                return 1;
 361        }
 362        return 0;
 363}
 364
 365static int grow_tree_refs(struct audit_context *ctx)
 366{
 367        struct audit_tree_refs *p = ctx->trees;
 368        ctx->trees = kzalloc(sizeof(struct audit_tree_refs), GFP_KERNEL);
 369        if (!ctx->trees) {
 370                ctx->trees = p;
 371                return 0;
 372        }
 373        if (p)
 374                p->next = ctx->trees;
 375        else
 376                ctx->first_trees = ctx->trees;
 377        ctx->tree_count = 31;
 378        return 1;
 379}
 380#endif
 381
 382static void unroll_tree_refs(struct audit_context *ctx,
 383                      struct audit_tree_refs *p, int count)
 384{
 385#ifdef CONFIG_AUDIT_TREE
 386        struct audit_tree_refs *q;
 387        int n;
 388        if (!p) {
 389                /* we started with empty chain */
 390                p = ctx->first_trees;
 391                count = 31;
 392                /* if the very first allocation has failed, nothing to do */
 393                if (!p)
 394                        return;
 395        }
 396        n = count;
 397        for (q = p; q != ctx->trees; q = q->next, n = 31) {
 398                while (n--) {
 399                        audit_put_chunk(q->c[n]);
 400                        q->c[n] = NULL;
 401                }
 402        }
 403        while (n-- > ctx->tree_count) {
 404                audit_put_chunk(q->c[n]);
 405                q->c[n] = NULL;
 406        }
 407        ctx->trees = p;
 408        ctx->tree_count = count;
 409#endif
 410}
 411
 412static void free_tree_refs(struct audit_context *ctx)
 413{
 414        struct audit_tree_refs *p, *q;
 415        for (p = ctx->first_trees; p; p = q) {
 416                q = p->next;
 417                kfree(p);
 418        }
 419}
 420
 421static int match_tree_refs(struct audit_context *ctx, struct audit_tree *tree)
 422{
 423#ifdef CONFIG_AUDIT_TREE
 424        struct audit_tree_refs *p;
 425        int n;
 426        if (!tree)
 427                return 0;
 428        /* full ones */
 429        for (p = ctx->first_trees; p != ctx->trees; p = p->next) {
 430                for (n = 0; n < 31; n++)
 431                        if (audit_tree_match(p->c[n], tree))
 432                                return 1;
 433        }
 434        /* partial */
 435        if (p) {
 436                for (n = ctx->tree_count; n < 31; n++)
 437                        if (audit_tree_match(p->c[n], tree))
 438                                return 1;
 439        }
 440#endif
 441        return 0;
 442}
 443
 444/* Determine if any context name data matches a rule's watch data */
 445/* Compare a task_struct with an audit_rule.  Return 1 on match, 0
 446 * otherwise.
 447 *
 448 * If task_creation is true, this is an explicit indication that we are
 449 * filtering a task rule at task creation time.  This and tsk == current are
 450 * the only situations where tsk->cred may be accessed without an rcu read lock.
 451 */
 452static int audit_filter_rules(struct task_struct *tsk,
 453                              struct audit_krule *rule,
 454                              struct audit_context *ctx,
 455                              struct audit_names *name,
 456                              enum audit_state *state,
 457                              bool task_creation)
 458{
 459        const struct cred *cred;
 460        int i, j, need_sid = 1;
 461        u32 sid;
 462
 463        cred = rcu_dereference_check(tsk->cred, tsk == current || task_creation);
 464
 465        for (i = 0; i < rule->field_count; i++) {
 466                struct audit_field *f = &rule->fields[i];
 467                int result = 0;
 468
 469                switch (f->type) {
 470                case AUDIT_PID:
 471                        result = audit_comparator(tsk->pid, f->op, f->val);
 472                        break;
 473                case AUDIT_PPID:
 474                        if (ctx) {
 475                                if (!ctx->ppid)
 476                                        ctx->ppid = sys_getppid();
 477                                result = audit_comparator(ctx->ppid, f->op, f->val);
 478                        }
 479                        break;
 480                case AUDIT_UID:
 481                        result = audit_comparator(cred->uid, f->op, f->val);
 482                        break;
 483                case AUDIT_EUID:
 484                        result = audit_comparator(cred->euid, f->op, f->val);
 485                        break;
 486                case AUDIT_SUID:
 487                        result = audit_comparator(cred->suid, f->op, f->val);
 488                        break;
 489                case AUDIT_FSUID:
 490                        result = audit_comparator(cred->fsuid, f->op, f->val);
 491                        break;
 492                case AUDIT_GID:
 493                        result = audit_comparator(cred->gid, f->op, f->val);
 494                        break;
 495                case AUDIT_EGID:
 496                        result = audit_comparator(cred->egid, f->op, f->val);
 497                        break;
 498                case AUDIT_SGID:
 499                        result = audit_comparator(cred->sgid, f->op, f->val);
 500                        break;
 501                case AUDIT_FSGID:
 502                        result = audit_comparator(cred->fsgid, f->op, f->val);
 503                        break;
 504                case AUDIT_PERS:
 505                        result = audit_comparator(tsk->personality, f->op, f->val);
 506                        break;
 507                case AUDIT_ARCH:
 508                        if (ctx)
 509                                result = audit_comparator(ctx->arch, f->op, f->val);
 510                        break;
 511
 512                case AUDIT_EXIT:
 513                        if (ctx && ctx->return_valid)
 514                                result = audit_comparator(ctx->return_code, f->op, f->val);
 515                        break;
 516                case AUDIT_SUCCESS:
 517                        if (ctx && ctx->return_valid) {
 518                                if (f->val)
 519                                        result = audit_comparator(ctx->return_valid, f->op, AUDITSC_SUCCESS);
 520                                else
 521                                        result = audit_comparator(ctx->return_valid, f->op, AUDITSC_FAILURE);
 522                        }
 523                        break;
 524                case AUDIT_DEVMAJOR:
 525                        if (name)
 526                                result = audit_comparator(MAJOR(name->dev),
 527                                                          f->op, f->val);
 528                        else if (ctx) {
 529                                for (j = 0; j < ctx->name_count; j++) {
 530                                        if (audit_comparator(MAJOR(ctx->names[j].dev),  f->op, f->val)) {
 531                                                ++result;
 532                                                break;
 533                                        }
 534                                }
 535                        }
 536                        break;
 537                case AUDIT_DEVMINOR:
 538                        if (name)
 539                                result = audit_comparator(MINOR(name->dev),
 540                                                          f->op, f->val);
 541                        else if (ctx) {
 542                                for (j = 0; j < ctx->name_count; j++) {
 543                                        if (audit_comparator(MINOR(ctx->names[j].dev), f->op, f->val)) {
 544                                                ++result;
 545                                                break;
 546                                        }
 547                                }
 548                        }
 549                        break;
 550                case AUDIT_INODE:
 551                        if (name)
 552                                result = (name->ino == f->val);
 553                        else if (ctx) {
 554                                for (j = 0; j < ctx->name_count; j++) {
 555                                        if (audit_comparator(ctx->names[j].ino, f->op, f->val)) {
 556                                                ++result;
 557                                                break;
 558                                        }
 559                                }
 560                        }
 561                        break;
 562                case AUDIT_WATCH:
 563                        if (name)
 564                                result = audit_watch_compare(rule->watch, name->ino, name->dev);
 565                        break;
 566                case AUDIT_DIR:
 567                        if (ctx)
 568                                result = match_tree_refs(ctx, rule->tree);
 569                        break;
 570                case AUDIT_LOGINUID:
 571                        result = 0;
 572                        if (ctx)
 573                                result = audit_comparator(tsk->loginuid, f->op, f->val);
 574                        break;
 575                case AUDIT_SUBJ_USER:
 576                case AUDIT_SUBJ_ROLE:
 577                case AUDIT_SUBJ_TYPE:
 578                case AUDIT_SUBJ_SEN:
 579                case AUDIT_SUBJ_CLR:
 580                        /* NOTE: this may return negative values indicating
 581                           a temporary error.  We simply treat this as a
 582                           match for now to avoid losing information that
 583                           may be wanted.   An error message will also be
 584                           logged upon error */
 585                        if (f->lsm_rule) {
 586                                if (need_sid) {
 587                                        security_task_getsecid(tsk, &sid);
 588                                        need_sid = 0;
 589                                }
 590                                result = security_audit_rule_match(sid, f->type,
 591                                                                  f->op,
 592                                                                  f->lsm_rule,
 593                                                                  ctx);
 594                        }
 595                        break;
 596                case AUDIT_OBJ_USER:
 597                case AUDIT_OBJ_ROLE:
 598                case AUDIT_OBJ_TYPE:
 599                case AUDIT_OBJ_LEV_LOW:
 600                case AUDIT_OBJ_LEV_HIGH:
 601                        /* The above note for AUDIT_SUBJ_USER...AUDIT_SUBJ_CLR
 602                           also applies here */
 603                        if (f->lsm_rule) {
 604                                /* Find files that match */
 605                                if (name) {
 606                                        result = security_audit_rule_match(
 607                                                   name->osid, f->type, f->op,
 608                                                   f->lsm_rule, ctx);
 609                                } else if (ctx) {
 610                                        for (j = 0; j < ctx->name_count; j++) {
 611                                                if (security_audit_rule_match(
 612                                                      ctx->names[j].osid,
 613                                                      f->type, f->op,
 614                                                      f->lsm_rule, ctx)) {
 615                                                        ++result;
 616                                                        break;
 617                                                }
 618                                        }
 619                                }
 620                                /* Find ipc objects that match */
 621                                if (!ctx || ctx->type != AUDIT_IPC)
 622                                        break;
 623                                if (security_audit_rule_match(ctx->ipc.osid,
 624                                                              f->type, f->op,
 625                                                              f->lsm_rule, ctx))
 626                                        ++result;
 627                        }
 628                        break;
 629                case AUDIT_ARG0:
 630                case AUDIT_ARG1:
 631                case AUDIT_ARG2:
 632                case AUDIT_ARG3:
 633                        if (ctx)
 634                                result = audit_comparator(ctx->argv[f->type-AUDIT_ARG0], f->op, f->val);
 635                        break;
 636                case AUDIT_FILTERKEY:
 637                        /* ignore this field for filtering */
 638                        result = 1;
 639                        break;
 640                case AUDIT_PERM:
 641                        result = audit_match_perm(ctx, f->val);
 642                        break;
 643                case AUDIT_FILETYPE:
 644                        result = audit_match_filetype(ctx, f->val);
 645                        break;
 646                }
 647
 648                if (!result)
 649                        return 0;
 650        }
 651
 652        if (ctx) {
 653                if (rule->prio <= ctx->prio)
 654                        return 0;
 655                if (rule->filterkey) {
 656                        kfree(ctx->filterkey);
 657                        ctx->filterkey = kstrdup(rule->filterkey, GFP_ATOMIC);
 658                }
 659                ctx->prio = rule->prio;
 660        }
 661        switch (rule->action) {
 662        case AUDIT_NEVER:    *state = AUDIT_DISABLED;       break;
 663        case AUDIT_ALWAYS:   *state = AUDIT_RECORD_CONTEXT; break;
 664        }
 665        return 1;
 666}
 667
 668/* At process creation time, we can determine if system-call auditing is
 669 * completely disabled for this task.  Since we only have the task
 670 * structure at this point, we can only check uid and gid.
 671 */
 672static enum audit_state audit_filter_task(struct task_struct *tsk, char **key)
 673{
 674        struct audit_entry *e;
 675        enum audit_state   state;
 676
 677        rcu_read_lock();
 678        list_for_each_entry_rcu(e, &audit_filter_list[AUDIT_FILTER_TASK], list) {
 679                if (audit_filter_rules(tsk, &e->rule, NULL, NULL,
 680                                       &state, true)) {
 681                        if (state == AUDIT_RECORD_CONTEXT)
 682                                *key = kstrdup(e->rule.filterkey, GFP_ATOMIC);
 683                        rcu_read_unlock();
 684                        return state;
 685                }
 686        }
 687        rcu_read_unlock();
 688        return AUDIT_BUILD_CONTEXT;
 689}
 690
 691/* At syscall entry and exit time, this filter is called if the
 692 * audit_state is not low enough that auditing cannot take place, but is
 693 * also not high enough that we already know we have to write an audit
 694 * record (i.e., the state is AUDIT_SETUP_CONTEXT or AUDIT_BUILD_CONTEXT).
 695 */
 696static enum audit_state audit_filter_syscall(struct task_struct *tsk,
 697                                             struct audit_context *ctx,
 698                                             struct list_head *list)
 699{
 700        struct audit_entry *e;
 701        enum audit_state state;
 702
 703        if (audit_pid && tsk->tgid == audit_pid)
 704                return AUDIT_DISABLED;
 705
 706        rcu_read_lock();
 707        if (!list_empty(list)) {
 708                int word = AUDIT_WORD(ctx->major);
 709                int bit  = AUDIT_BIT(ctx->major);
 710
 711                list_for_each_entry_rcu(e, list, list) {
 712                        if ((e->rule.mask[word] & bit) == bit &&
 713                            audit_filter_rules(tsk, &e->rule, ctx, NULL,
 714                                               &state, false)) {
 715                                rcu_read_unlock();
 716                                ctx->current_state = state;
 717                                return state;
 718                        }
 719                }
 720        }
 721        rcu_read_unlock();
 722        return AUDIT_BUILD_CONTEXT;
 723}
 724
 725/* At syscall exit time, this filter is called if any audit_names[] have been
 726 * collected during syscall processing.  We only check rules in sublists at hash
 727 * buckets applicable to the inode numbers in audit_names[].
 728 * Regarding audit_state, same rules apply as for audit_filter_syscall().
 729 */
 730void audit_filter_inodes(struct task_struct *tsk, struct audit_context *ctx)
 731{
 732        int i;
 733        struct audit_entry *e;
 734        enum audit_state state;
 735
 736        if (audit_pid && tsk->tgid == audit_pid)
 737                return;
 738
 739        rcu_read_lock();
 740        for (i = 0; i < ctx->name_count; i++) {
 741                int word = AUDIT_WORD(ctx->major);
 742                int bit  = AUDIT_BIT(ctx->major);
 743                struct audit_names *n = &ctx->names[i];
 744                int h = audit_hash_ino((u32)n->ino);
 745                struct list_head *list = &audit_inode_hash[h];
 746
 747                if (list_empty(list))
 748                        continue;
 749
 750                list_for_each_entry_rcu(e, list, list) {
 751                        if ((e->rule.mask[word] & bit) == bit &&
 752                            audit_filter_rules(tsk, &e->rule, ctx, n,
 753                                               &state, false)) {
 754                                rcu_read_unlock();
 755                                ctx->current_state = state;
 756                                return;
 757                        }
 758                }
 759        }
 760        rcu_read_unlock();
 761}
 762
 763static inline struct audit_context *audit_get_context(struct task_struct *tsk,
 764                                                      int return_valid,
 765                                                      long return_code)
 766{
 767        struct audit_context *context = tsk->audit_context;
 768
 769        if (likely(!context))
 770                return NULL;
 771        context->return_valid = return_valid;
 772
 773        /*
 774         * we need to fix up the return code in the audit logs if the actual
 775         * return codes are later going to be fixed up by the arch specific
 776         * signal handlers
 777         *
 778         * This is actually a test for:
 779         * (rc == ERESTARTSYS ) || (rc == ERESTARTNOINTR) ||
 780         * (rc == ERESTARTNOHAND) || (rc == ERESTART_RESTARTBLOCK)
 781         *
 782         * but is faster than a bunch of ||
 783         */
 784        if (unlikely(return_code <= -ERESTARTSYS) &&
 785            (return_code >= -ERESTART_RESTARTBLOCK) &&
 786            (return_code != -ENOIOCTLCMD))
 787                context->return_code = -EINTR;
 788        else
 789                context->return_code  = return_code;
 790
 791        if (context->in_syscall && !context->dummy) {
 792                audit_filter_syscall(tsk, context, &audit_filter_list[AUDIT_FILTER_EXIT]);
 793                audit_filter_inodes(tsk, context);
 794        }
 795
 796        tsk->audit_context = NULL;
 797        return context;
 798}
 799
 800static inline void audit_free_names(struct audit_context *context)
 801{
 802        int i;
 803
 804#if AUDIT_DEBUG == 2
 805        if (context->put_count + context->ino_count != context->name_count) {
 806                printk(KERN_ERR "%s:%d(:%d): major=%d in_syscall=%d"
 807                       " name_count=%d put_count=%d"
 808                       " ino_count=%d [NOT freeing]\n",
 809                       __FILE__, __LINE__,
 810                       context->serial, context->major, context->in_syscall,
 811                       context->name_count, context->put_count,
 812                       context->ino_count);
 813                for (i = 0; i < context->name_count; i++) {
 814                        printk(KERN_ERR "names[%d] = %p = %s\n", i,
 815                               context->names[i].name,
 816                               context->names[i].name ?: "(null)");
 817                }
 818                dump_stack();
 819                return;
 820        }
 821#endif
 822#if AUDIT_DEBUG
 823        context->put_count  = 0;
 824        context->ino_count  = 0;
 825#endif
 826
 827        for (i = 0; i < context->name_count; i++) {
 828                if (context->names[i].name && context->names[i].name_put)
 829                        __putname(context->names[i].name);
 830        }
 831        context->name_count = 0;
 832        path_put(&context->pwd);
 833        context->pwd.dentry = NULL;
 834        context->pwd.mnt = NULL;
 835}
 836
 837static inline void audit_free_aux(struct audit_context *context)
 838{
 839        struct audit_aux_data *aux;
 840
 841        while ((aux = context->aux)) {
 842                context->aux = aux->next;
 843                kfree(aux);
 844        }
 845        while ((aux = context->aux_pids)) {
 846                context->aux_pids = aux->next;
 847                kfree(aux);
 848        }
 849}
 850
 851static inline void audit_zero_context(struct audit_context *context,
 852                                      enum audit_state state)
 853{
 854        memset(context, 0, sizeof(*context));
 855        context->state      = state;
 856        context->prio = state == AUDIT_RECORD_CONTEXT ? ~0ULL : 0;
 857}
 858
 859static inline struct audit_context *audit_alloc_context(enum audit_state state)
 860{
 861        struct audit_context *context;
 862
 863        if (!(context = kmalloc(sizeof(*context), GFP_KERNEL)))
 864                return NULL;
 865        audit_zero_context(context, state);
 866        INIT_LIST_HEAD(&context->killed_trees);
 867        return context;
 868}
 869
 870/**
 871 * audit_alloc - allocate an audit context block for a task
 872 * @tsk: task
 873 *
 874 * Filter on the task information and allocate a per-task audit context
 875 * if necessary.  Doing so turns on system call auditing for the
 876 * specified task.  This is called from copy_process, so no lock is
 877 * needed.
 878 */
 879int audit_alloc(struct task_struct *tsk)
 880{
 881        struct audit_context *context;
 882        enum audit_state     state;
 883        char *key = NULL;
 884
 885        if (likely(!audit_ever_enabled))
 886                return 0; /* Return if not auditing. */
 887
 888        state = audit_filter_task(tsk, &key);
 889        if (likely(state == AUDIT_DISABLED))
 890                return 0;
 891
 892        if (!(context = audit_alloc_context(state))) {
 893                kfree(key);
 894                audit_log_lost("out of memory in audit_alloc");
 895                return -ENOMEM;
 896        }
 897        context->filterkey = key;
 898
 899        tsk->audit_context  = context;
 900        set_tsk_thread_flag(tsk, TIF_SYSCALL_AUDIT);
 901        return 0;
 902}
 903
 904static inline void audit_free_context(struct audit_context *context)
 905{
 906        struct audit_context *previous;
 907        int                  count = 0;
 908
 909        do {
 910                previous = context->previous;
 911                if (previous || (count &&  count < 10)) {
 912                        ++count;
 913                        printk(KERN_ERR "audit(:%d): major=%d name_count=%d:"
 914                               " freeing multiple contexts (%d)\n",
 915                               context->serial, context->major,
 916                               context->name_count, count);
 917                }
 918                audit_free_names(context);
 919                unroll_tree_refs(context, NULL, 0);
 920                free_tree_refs(context);
 921                audit_free_aux(context);
 922                kfree(context->filterkey);
 923                kfree(context->sockaddr);
 924                kfree(context);
 925                context  = previous;
 926        } while (context);
 927        if (count >= 10)
 928                printk(KERN_ERR "audit: freed %d contexts\n", count);
 929}
 930
 931void audit_log_task_context(struct audit_buffer *ab)
 932{
 933        char *ctx = NULL;
 934        unsigned len;
 935        int error;
 936        u32 sid;
 937
 938        security_task_getsecid(current, &sid);
 939        if (!sid)
 940                return;
 941
 942        error = security_secid_to_secctx(sid, &ctx, &len);
 943        if (error) {
 944                if (error != -EINVAL)
 945                        goto error_path;
 946                return;
 947        }
 948
 949        audit_log_format(ab, " subj=%s", ctx);
 950        security_release_secctx(ctx, len);
 951        return;
 952
 953error_path:
 954        audit_panic("error in audit_log_task_context");
 955        return;
 956}
 957
 958EXPORT_SYMBOL(audit_log_task_context);
 959
 960static void audit_log_task_info(struct audit_buffer *ab, struct task_struct *tsk)
 961{
 962        char name[sizeof(tsk->comm)];
 963        struct mm_struct *mm = tsk->mm;
 964        struct vm_area_struct *vma;
 965
 966        /* tsk == current */
 967
 968        get_task_comm(name, tsk);
 969        audit_log_format(ab, " comm=");
 970        audit_log_untrustedstring(ab, name);
 971
 972        if (mm) {
 973                down_read(&mm->mmap_sem);
 974                vma = mm->mmap;
 975                while (vma) {
 976                        if ((vma->vm_flags & VM_EXECUTABLE) &&
 977                            vma->vm_file) {
 978                                audit_log_d_path(ab, "exe=",
 979                                                 &vma->vm_file->f_path);
 980                                break;
 981                        }
 982                        vma = vma->vm_next;
 983                }
 984                up_read(&mm->mmap_sem);
 985        }
 986        audit_log_task_context(ab);
 987}
 988
 989static int audit_log_pid_context(struct audit_context *context, pid_t pid,
 990                                 uid_t auid, uid_t uid, unsigned int sessionid,
 991                                 u32 sid, char *comm)
 992{
 993        struct audit_buffer *ab;
 994        char *ctx = NULL;
 995        u32 len;
 996        int rc = 0;
 997
 998        ab = audit_log_start(context, GFP_KERNEL, AUDIT_OBJ_PID);
 999        if (!ab)
1000                return rc;
1001
1002        audit_log_format(ab, "opid=%d oauid=%d ouid=%d oses=%d", pid, auid,
1003                         uid, sessionid);
1004        if (security_secid_to_secctx(sid, &ctx, &len)) {
1005                audit_log_format(ab, " obj=(none)");
1006                rc = 1;
1007        } else {
1008                audit_log_format(ab, " obj=%s", ctx);
1009                security_release_secctx(ctx, len);
1010        }
1011        audit_log_format(ab, " ocomm=");
1012        audit_log_untrustedstring(ab, comm);
1013        audit_log_end(ab);
1014
1015        return rc;
1016}
1017
1018/*
1019 * to_send and len_sent accounting are very loose estimates.  We aren't
1020 * really worried about a hard cap to MAX_EXECVE_AUDIT_LEN so much as being
1021 * within about 500 bytes (next page boundary)
1022 *
1023 * why snprintf?  an int is up to 12 digits long.  if we just assumed when
1024 * logging that a[%d]= was going to be 16 characters long we would be wasting
1025 * space in every audit message.  In one 7500 byte message we can log up to
1026 * about 1000 min size arguments.  That comes down to about 50% waste of space
1027 * if we didn't do the snprintf to find out how long arg_num_len was.
1028 */
1029static int audit_log_single_execve_arg(struct audit_context *context,
1030                                        struct audit_buffer **ab,
1031                                        int arg_num,
1032                                        size_t *len_sent,
1033                                        const char __user *p,
1034                                        char *buf)
1035{
1036        char arg_num_len_buf[12];
1037        const char __user *tmp_p = p;
1038        /* how many digits are in arg_num? 5 is the length of ' a=""' */
1039        size_t arg_num_len = snprintf(arg_num_len_buf, 12, "%d", arg_num) + 5;
1040        size_t len, len_left, to_send;
1041        size_t max_execve_audit_len = MAX_EXECVE_AUDIT_LEN;
1042        unsigned int i, has_cntl = 0, too_long = 0;
1043        int ret;
1044
1045        /* strnlen_user includes the null we don't want to send */
1046        len_left = len = strnlen_user(p, MAX_ARG_STRLEN) - 1;
1047
1048        /*
1049         * We just created this mm, if we can't find the strings
1050         * we just copied into it something is _very_ wrong. Similar
1051         * for strings that are too long, we should not have created
1052         * any.
1053         */
1054        if (unlikely((len == -1) || len > MAX_ARG_STRLEN - 1)) {
1055                WARN_ON(1);
1056                send_sig(SIGKILL, current, 0);
1057                return -1;
1058        }
1059
1060        /* walk the whole argument looking for non-ascii chars */
1061        do {
1062                if (len_left > MAX_EXECVE_AUDIT_LEN)
1063                        to_send = MAX_EXECVE_AUDIT_LEN;
1064                else
1065                        to_send = len_left;
1066                ret = copy_from_user(buf, tmp_p, to_send);
1067                /*
1068                 * There is no reason for this copy to be short. We just
1069                 * copied them here, and the mm hasn't been exposed to user-
1070                 * space yet.
1071                 */
1072                if (ret) {
1073                        WARN_ON(1);
1074                        send_sig(SIGKILL, current, 0);
1075                        return -1;
1076                }
1077                buf[to_send] = '\0';
1078                has_cntl = audit_string_contains_control(buf, to_send);
1079                if (has_cntl) {
1080                        /*
1081                         * hex messages get logged as 2 bytes, so we can only
1082                         * send half as much in each message
1083                         */
1084                        max_execve_audit_len = MAX_EXECVE_AUDIT_LEN / 2;
1085                        break;
1086                }
1087                len_left -= to_send;
1088                tmp_p += to_send;
1089        } while (len_left > 0);
1090
1091        len_left = len;
1092
1093        if (len > max_execve_audit_len)
1094                too_long = 1;
1095
1096        /* rewalk the argument actually logging the message */
1097        for (i = 0; len_left > 0; i++) {
1098                int room_left;
1099
1100                if (len_left > max_execve_audit_len)
1101                        to_send = max_execve_audit_len;
1102                else
1103                        to_send = len_left;
1104
1105                /* do we have space left to send this argument in this ab? */
1106                room_left = MAX_EXECVE_AUDIT_LEN - arg_num_len - *len_sent;
1107                if (has_cntl)
1108                        room_left -= (to_send * 2);
1109                else
1110                        room_left -= to_send;
1111                if (room_left < 0) {
1112                        *len_sent = 0;
1113                        audit_log_end(*ab);
1114                        *ab = audit_log_start(context, GFP_KERNEL, AUDIT_EXECVE);
1115                        if (!*ab)
1116                                return 0;
1117                }
1118
1119                /*
1120                 * first record needs to say how long the original string was
1121                 * so we can be sure nothing was lost.
1122                 */
1123                if ((i == 0) && (too_long))
1124                        audit_log_format(*ab, " a%d_len=%zu", arg_num,
1125                                         has_cntl ? 2*len : len);
1126
1127                /*
1128                 * normally arguments are small enough to fit and we already
1129                 * filled buf above when we checked for control characters
1130                 * so don't bother with another copy_from_user
1131                 */
1132                if (len >= max_execve_audit_len)
1133                        ret = copy_from_user(buf, p, to_send);
1134                else
1135                        ret = 0;
1136                if (ret) {
1137                        WARN_ON(1);
1138                        send_sig(SIGKILL, current, 0);
1139                        return -1;
1140                }
1141                buf[to_send] = '\0';
1142
1143                /* actually log it */
1144                audit_log_format(*ab, " a%d", arg_num);
1145                if (too_long)
1146                        audit_log_format(*ab, "[%d]", i);
1147                audit_log_format(*ab, "=");
1148                if (has_cntl)
1149                        audit_log_n_hex(*ab, buf, to_send);
1150                else
1151                        audit_log_string(*ab, buf);
1152
1153                p += to_send;
1154                len_left -= to_send;
1155                *len_sent += arg_num_len;
1156                if (has_cntl)
1157                        *len_sent += to_send * 2;
1158                else
1159                        *len_sent += to_send;
1160        }
1161        /* include the null we didn't log */
1162        return len + 1;
1163}
1164
1165static void audit_log_execve_info(struct audit_context *context,
1166                                  struct audit_buffer **ab,
1167                                  struct audit_aux_data_execve *axi)
1168{
1169        int i;
1170        size_t len, len_sent = 0;
1171        const char __user *p;
1172        char *buf;
1173
1174        if (axi->mm != current->mm)
1175                return; /* execve failed, no additional info */
1176
1177        p = (const char __user *)axi->mm->arg_start;
1178
1179        audit_log_format(*ab, "argc=%d", axi->argc);
1180
1181        /*
1182         * we need some kernel buffer to hold the userspace args.  Just
1183         * allocate one big one rather than allocating one of the right size
1184         * for every single argument inside audit_log_single_execve_arg()
1185         * should be <8k allocation so should be pretty safe.
1186         */
1187        buf = kmalloc(MAX_EXECVE_AUDIT_LEN + 1, GFP_KERNEL);
1188        if (!buf) {
1189                audit_panic("out of memory for argv string\n");
1190                return;
1191        }
1192
1193        for (i = 0; i < axi->argc; i++) {
1194                len = audit_log_single_execve_arg(context, ab, i,
1195                                                  &len_sent, p, buf);
1196                if (len <= 0)
1197                        break;
1198                p += len;
1199        }
1200        kfree(buf);
1201}
1202
1203static void audit_log_cap(struct audit_buffer *ab, char *prefix, kernel_cap_t *cap)
1204{
1205        int i;
1206
1207        audit_log_format(ab, " %s=", prefix);
1208        CAP_FOR_EACH_U32(i) {
1209                audit_log_format(ab, "%08x", cap->cap[(_KERNEL_CAPABILITY_U32S-1) - i]);
1210        }
1211}
1212
1213static void audit_log_fcaps(struct audit_buffer *ab, struct audit_names *name)
1214{
1215        kernel_cap_t *perm = &name->fcap.permitted;
1216        kernel_cap_t *inh = &name->fcap.inheritable;
1217        int log = 0;
1218
1219        if (!cap_isclear(*perm)) {
1220                audit_log_cap(ab, "cap_fp", perm);
1221                log = 1;
1222        }
1223        if (!cap_isclear(*inh)) {
1224                audit_log_cap(ab, "cap_fi", inh);
1225                log = 1;
1226        }
1227
1228        if (log)
1229                audit_log_format(ab, " cap_fe=%d cap_fver=%x", name->fcap.fE, name->fcap_ver);
1230}
1231
1232static void show_special(struct audit_context *context, int *call_panic)
1233{
1234        struct audit_buffer *ab;
1235        int i;
1236
1237        ab = audit_log_start(context, GFP_KERNEL, context->type);
1238        if (!ab)
1239                return;
1240
1241        switch (context->type) {
1242        case AUDIT_SOCKETCALL: {
1243                int nargs = context->socketcall.nargs;
1244                audit_log_format(ab, "nargs=%d", nargs);
1245                for (i = 0; i < nargs; i++)
1246                        audit_log_format(ab, " a%d=%lx", i,
1247                                context->socketcall.args[i]);
1248                break; }
1249        case AUDIT_IPC: {
1250                u32 osid = context->ipc.osid;
1251
1252                audit_log_format(ab, "ouid=%u ogid=%u mode=%#o",
1253                         context->ipc.uid, context->ipc.gid, context->ipc.mode);
1254                if (osid) {
1255                        char *ctx = NULL;
1256                        u32 len;
1257                        if (security_secid_to_secctx(osid, &ctx, &len)) {
1258                                audit_log_format(ab, " osid=%u", osid);
1259                                *call_panic = 1;
1260                        } else {
1261                                audit_log_format(ab, " obj=%s", ctx);
1262                                security_release_secctx(ctx, len);
1263                        }
1264                }
1265                if (context->ipc.has_perm) {
1266                        audit_log_end(ab);
1267                        ab = audit_log_start(context, GFP_KERNEL,
1268                                             AUDIT_IPC_SET_PERM);
1269                        audit_log_format(ab,
1270                                "qbytes=%lx ouid=%u ogid=%u mode=%#o",
1271                                context->ipc.qbytes,
1272                                context->ipc.perm_uid,
1273                                context->ipc.perm_gid,
1274                                context->ipc.perm_mode);
1275                        if (!ab)
1276                                return;
1277                }
1278                break; }
1279        case AUDIT_MQ_OPEN: {
1280                audit_log_format(ab,
1281                        "oflag=0x%x mode=%#o mq_flags=0x%lx mq_maxmsg=%ld "
1282                        "mq_msgsize=%ld mq_curmsgs=%ld",
1283                        context->mq_open.oflag, context->mq_open.mode,
1284                        context->mq_open.attr.mq_flags,
1285                        context->mq_open.attr.mq_maxmsg,
1286                        context->mq_open.attr.mq_msgsize,
1287                        context->mq_open.attr.mq_curmsgs);
1288                break; }
1289        case AUDIT_MQ_SENDRECV: {
1290                audit_log_format(ab,
1291                        "mqdes=%d msg_len=%zd msg_prio=%u "
1292                        "abs_timeout_sec=%ld abs_timeout_nsec=%ld",
1293                        context->mq_sendrecv.mqdes,
1294                        context->mq_sendrecv.msg_len,
1295                        context->mq_sendrecv.msg_prio,
1296                        context->mq_sendrecv.abs_timeout.tv_sec,
1297                        context->mq_sendrecv.abs_timeout.tv_nsec);
1298                break; }
1299        case AUDIT_MQ_NOTIFY: {
1300                audit_log_format(ab, "mqdes=%d sigev_signo=%d",
1301                                context->mq_notify.mqdes,
1302                                context->mq_notify.sigev_signo);
1303                break; }
1304        case AUDIT_MQ_GETSETATTR: {
1305                struct mq_attr *attr = &context->mq_getsetattr.mqstat;
1306                audit_log_format(ab,
1307                        "mqdes=%d mq_flags=0x%lx mq_maxmsg=%ld mq_msgsize=%ld "
1308                        "mq_curmsgs=%ld ",
1309                        context->mq_getsetattr.mqdes,
1310                        attr->mq_flags, attr->mq_maxmsg,
1311                        attr->mq_msgsize, attr->mq_curmsgs);
1312                break; }
1313        case AUDIT_CAPSET: {
1314                audit_log_format(ab, "pid=%d", context->capset.pid);
1315                audit_log_cap(ab, "cap_pi", &context->capset.cap.inheritable);
1316                audit_log_cap(ab, "cap_pp", &context->capset.cap.permitted);
1317                audit_log_cap(ab, "cap_pe", &context->capset.cap.effective);
1318                break; }
1319        case AUDIT_MMAP: {
1320                audit_log_format(ab, "fd=%d flags=0x%x", context->mmap.fd,
1321                                 context->mmap.flags);
1322                break; }
1323        }
1324        audit_log_end(ab);
1325}
1326
1327static void audit_log_exit(struct audit_context *context, struct task_struct *tsk)
1328{
1329        const struct cred *cred;
1330        int i, call_panic = 0;
1331        struct audit_buffer *ab;
1332        struct audit_aux_data *aux;
1333        const char *tty;
1334
1335        /* tsk == current */
1336        context->pid = tsk->pid;
1337        if (!context->ppid)
1338                context->ppid = sys_getppid();
1339        cred = current_cred();
1340        context->uid   = cred->uid;
1341        context->gid   = cred->gid;
1342        context->euid  = cred->euid;
1343        context->suid  = cred->suid;
1344        context->fsuid = cred->fsuid;
1345        context->egid  = cred->egid;
1346        context->sgid  = cred->sgid;
1347        context->fsgid = cred->fsgid;
1348        context->personality = tsk->personality;
1349
1350        ab = audit_log_start(context, GFP_KERNEL, AUDIT_SYSCALL);
1351        if (!ab)
1352                return;         /* audit_panic has been called */
1353        audit_log_format(ab, "arch=%x syscall=%d",
1354                         context->arch, context->major);
1355        if (context->personality != PER_LINUX)
1356                audit_log_format(ab, " per=%lx", context->personality);
1357        if (context->return_valid)
1358                audit_log_format(ab, " success=%s exit=%ld",
1359                                 (context->return_valid==AUDITSC_SUCCESS)?"yes":"no",
1360                                 context->return_code);
1361
1362        spin_lock_irq(&tsk->sighand->siglock);
1363        if (tsk->signal && tsk->signal->tty && tsk->signal->tty->name)
1364                tty = tsk->signal->tty->name;
1365        else
1366                tty = "(none)";
1367        spin_unlock_irq(&tsk->sighand->siglock);
1368
1369        audit_log_format(ab,
1370                  " a0=%lx a1=%lx a2=%lx a3=%lx items=%d"
1371                  " ppid=%d pid=%d auid=%u uid=%u gid=%u"
1372                  " euid=%u suid=%u fsuid=%u"
1373                  " egid=%u sgid=%u fsgid=%u tty=%s ses=%u",
1374                  context->argv[0],
1375                  context->argv[1],
1376                  context->argv[2],
1377                  context->argv[3],
1378                  context->name_count,
1379                  context->ppid,
1380                  context->pid,
1381                  tsk->loginuid,
1382                  context->uid,
1383                  context->gid,
1384                  context->euid, context->suid, context->fsuid,
1385                  context->egid, context->sgid, context->fsgid, tty,
1386                  tsk->sessionid);
1387
1388
1389        audit_log_task_info(ab, tsk);
1390        audit_log_key(ab, context->filterkey);
1391        audit_log_end(ab);
1392
1393        for (aux = context->aux; aux; aux = aux->next) {
1394
1395                ab = audit_log_start(context, GFP_KERNEL, aux->type);
1396                if (!ab)
1397                        continue; /* audit_panic has been called */
1398
1399                switch (aux->type) {
1400
1401                case AUDIT_EXECVE: {
1402                        struct audit_aux_data_execve *axi = (void *)aux;
1403                        audit_log_execve_info(context, &ab, axi);
1404                        break; }
1405
1406                case AUDIT_BPRM_FCAPS: {
1407                        struct audit_aux_data_bprm_fcaps *axs = (void *)aux;
1408                        audit_log_format(ab, "fver=%x", axs->fcap_ver);
1409                        audit_log_cap(ab, "fp", &axs->fcap.permitted);
1410                        audit_log_cap(ab, "fi", &axs->fcap.inheritable);
1411                        audit_log_format(ab, " fe=%d", axs->fcap.fE);
1412                        audit_log_cap(ab, "old_pp", &axs->old_pcap.permitted);
1413                        audit_log_cap(ab, "old_pi", &axs->old_pcap.inheritable);
1414                        audit_log_cap(ab, "old_pe", &axs->old_pcap.effective);
1415                        audit_log_cap(ab, "new_pp", &axs->new_pcap.permitted);
1416                        audit_log_cap(ab, "new_pi", &axs->new_pcap.inheritable);
1417                        audit_log_cap(ab, "new_pe", &axs->new_pcap.effective);
1418                        break; }
1419
1420                }
1421                audit_log_end(ab);
1422        }
1423
1424        if (context->type)
1425                show_special(context, &call_panic);
1426
1427        if (context->fds[0] >= 0) {
1428                ab = audit_log_start(context, GFP_KERNEL, AUDIT_FD_PAIR);
1429                if (ab) {
1430                        audit_log_format(ab, "fd0=%d fd1=%d",
1431                                        context->fds[0], context->fds[1]);
1432                        audit_log_end(ab);
1433                }
1434        }
1435
1436        if (context->sockaddr_len) {
1437                ab = audit_log_start(context, GFP_KERNEL, AUDIT_SOCKADDR);
1438                if (ab) {
1439                        audit_log_format(ab, "saddr=");
1440                        audit_log_n_hex(ab, (void *)context->sockaddr,
1441                                        context->sockaddr_len);
1442                        audit_log_end(ab);
1443                }
1444        }
1445
1446        for (aux = context->aux_pids; aux; aux = aux->next) {
1447                struct audit_aux_data_pids *axs = (void *)aux;
1448
1449                for (i = 0; i < axs->pid_count; i++)
1450                        if (audit_log_pid_context(context, axs->target_pid[i],
1451                                                  axs->target_auid[i],
1452                                                  axs->target_uid[i],
1453                                                  axs->target_sessionid[i],
1454                                                  axs->target_sid[i],
1455                                                  axs->target_comm[i]))
1456                                call_panic = 1;
1457        }
1458
1459        if (context->target_pid &&
1460            audit_log_pid_context(context, context->target_pid,
1461                                  context->target_auid, context->target_uid,
1462                                  context->target_sessionid,
1463                                  context->target_sid, context->target_comm))
1464                        call_panic = 1;
1465
1466        if (context->pwd.dentry && context->pwd.mnt) {
1467                ab = audit_log_start(context, GFP_KERNEL, AUDIT_CWD);
1468                if (ab) {
1469                        audit_log_d_path(ab, "cwd=", &context->pwd);
1470                        audit_log_end(ab);
1471                }
1472        }
1473        for (i = 0; i < context->name_count; i++) {
1474                struct audit_names *n = &context->names[i];
1475
1476                ab = audit_log_start(context, GFP_KERNEL, AUDIT_PATH);
1477                if (!ab)
1478                        continue; /* audit_panic has been called */
1479
1480                audit_log_format(ab, "item=%d", i);
1481
1482                if (n->name) {
1483                        switch(n->name_len) {
1484                        case AUDIT_NAME_FULL:
1485                                /* log the full path */
1486                                audit_log_format(ab, " name=");
1487                                audit_log_untrustedstring(ab, n->name);
1488                                break;
1489                        case 0:
1490                                /* name was specified as a relative path and the
1491                                 * directory component is the cwd */
1492                                audit_log_d_path(ab, "name=", &context->pwd);
1493                                break;
1494                        default:
1495                                /* log the name's directory component */
1496                                audit_log_format(ab, " name=");
1497                                audit_log_n_untrustedstring(ab, n->name,
1498                                                            n->name_len);
1499                        }
1500                } else
1501                        audit_log_format(ab, " name=(null)");
1502
1503                if (n->ino != (unsigned long)-1) {
1504                        audit_log_format(ab, " inode=%lu"
1505                                         " dev=%02x:%02x mode=%#o"
1506                                         " ouid=%u ogid=%u rdev=%02x:%02x",
1507                                         n->ino,
1508                                         MAJOR(n->dev),
1509                                         MINOR(n->dev),
1510                                         n->mode,
1511                                         n->uid,
1512                                         n->gid,
1513                                         MAJOR(n->rdev),
1514                                         MINOR(n->rdev));
1515                }
1516                if (n->osid != 0) {
1517                        char *ctx = NULL;
1518                        u32 len;
1519                        if (security_secid_to_secctx(
1520                                n->osid, &ctx, &len)) {
1521                                audit_log_format(ab, " osid=%u", n->osid);
1522                                call_panic = 2;
1523                        } else {
1524                                audit_log_format(ab, " obj=%s", ctx);
1525                                security_release_secctx(ctx, len);
1526                        }
1527                }
1528
1529                audit_log_fcaps(ab, n);
1530
1531                audit_log_end(ab);
1532        }
1533
1534        /* Send end of event record to help user space know we are finished */
1535        ab = audit_log_start(context, GFP_KERNEL, AUDIT_EOE);
1536        if (ab)
1537                audit_log_end(ab);
1538        if (call_panic)
1539                audit_panic("error converting sid to string");
1540}
1541
1542/**
1543 * audit_free - free a per-task audit context
1544 * @tsk: task whose audit context block to free
1545 *
1546 * Called from copy_process and do_exit
1547 */
1548void audit_free(struct task_struct *tsk)
1549{
1550        struct audit_context *context;
1551
1552        context = audit_get_context(tsk, 0, 0);
1553        if (likely(!context))
1554                return;
1555
1556        /* Check for system calls that do not go through the exit
1557         * function (e.g., exit_group), then free context block.
1558         * We use GFP_ATOMIC here because we might be doing this
1559         * in the context of the idle thread */
1560        /* that can happen only if we are called from do_exit() */
1561        if (context->in_syscall && context->current_state == AUDIT_RECORD_CONTEXT)
1562                audit_log_exit(context, tsk);
1563        if (!list_empty(&context->killed_trees))
1564                audit_kill_trees(&context->killed_trees);
1565
1566        audit_free_context(context);
1567}
1568
1569/**
1570 * audit_syscall_entry - fill in an audit record at syscall entry
1571 * @arch: architecture type
1572 * @major: major syscall type (function)
1573 * @a1: additional syscall register 1
1574 * @a2: additional syscall register 2
1575 * @a3: additional syscall register 3
1576 * @a4: additional syscall register 4
1577 *
1578 * Fill in audit context at syscall entry.  This only happens if the
1579 * audit context was created when the task was created and the state or
1580 * filters demand the audit context be built.  If the state from the
1581 * per-task filter or from the per-syscall filter is AUDIT_RECORD_CONTEXT,
1582 * then the record will be written at syscall exit time (otherwise, it
1583 * will only be written if another part of the kernel requests that it
1584 * be written).
1585 */
1586void audit_syscall_entry(int arch, int major,
1587                         unsigned long a1, unsigned long a2,
1588                         unsigned long a3, unsigned long a4)
1589{
1590        struct task_struct *tsk = current;
1591        struct audit_context *context = tsk->audit_context;
1592        enum audit_state     state;
1593
1594        if (unlikely(!context))
1595                return;
1596
1597        /*
1598         * This happens only on certain architectures that make system
1599         * calls in kernel_thread via the entry.S interface, instead of
1600         * with direct calls.  (If you are porting to a new
1601         * architecture, hitting this condition can indicate that you
1602         * got the _exit/_leave calls backward in entry.S.)
1603         *
1604         * i386     no
1605         * x86_64   no
1606         * ppc64    yes (see arch/powerpc/platforms/iseries/misc.S)
1607         *
1608         * This also happens with vm86 emulation in a non-nested manner
1609         * (entries without exits), so this case must be caught.
1610         */
1611        if (context->in_syscall) {
1612                struct audit_context *newctx;
1613
1614#if AUDIT_DEBUG
1615                printk(KERN_ERR
1616                       "audit(:%d) pid=%d in syscall=%d;"
1617                       " entering syscall=%d\n",
1618                       context->serial, tsk->pid, context->major, major);
1619#endif
1620                newctx = audit_alloc_context(context->state);
1621                if (newctx) {
1622                        newctx->previous   = context;
1623                        context            = newctx;
1624                        tsk->audit_context = newctx;
1625                } else  {
1626                        /* If we can't alloc a new context, the best we
1627                         * can do is to leak memory (any pending putname
1628                         * will be lost).  The only other alternative is
1629                         * to abandon auditing. */
1630                        audit_zero_context(context, context->state);
1631                }
1632        }
1633        BUG_ON(context->in_syscall || context->name_count);
1634
1635        if (!audit_enabled)
1636                return;
1637
1638        context->arch       = arch;
1639        context->major      = major;
1640        context->argv[0]    = a1;
1641        context->argv[1]    = a2;
1642        context->argv[2]    = a3;
1643        context->argv[3]    = a4;
1644
1645        state = context->state;
1646        context->dummy = !audit_n_rules;
1647        if (!context->dummy && state == AUDIT_BUILD_CONTEXT) {
1648                context->prio = 0;
1649                state = audit_filter_syscall(tsk, context, &audit_filter_list[AUDIT_FILTER_ENTRY]);
1650        }
1651        if (likely(state == AUDIT_DISABLED))
1652                return;
1653
1654        context->serial     = 0;
1655        context->ctime      = CURRENT_TIME;
1656        context->in_syscall = 1;
1657        context->current_state  = state;
1658        context->ppid       = 0;
1659}
1660
1661void audit_finish_fork(struct task_struct *child)
1662{
1663        struct audit_context *ctx = current->audit_context;
1664        struct audit_context *p = child->audit_context;
1665        if (!p || !ctx)
1666                return;
1667        if (!ctx->in_syscall || ctx->current_state != AUDIT_RECORD_CONTEXT)
1668                return;
1669        p->arch = ctx->arch;
1670        p->major = ctx->major;
1671        memcpy(p->argv, ctx->argv, sizeof(ctx->argv));
1672        p->ctime = ctx->ctime;
1673        p->dummy = ctx->dummy;
1674        p->in_syscall = ctx->in_syscall;
1675        p->filterkey = kstrdup(ctx->filterkey, GFP_KERNEL);
1676        p->ppid = current->pid;
1677        p->prio = ctx->prio;
1678        p->current_state = ctx->current_state;
1679}
1680
1681/**
1682 * audit_syscall_exit - deallocate audit context after a system call
1683 * @valid: success/failure flag
1684 * @return_code: syscall return value
1685 *
1686 * Tear down after system call.  If the audit context has been marked as
1687 * auditable (either because of the AUDIT_RECORD_CONTEXT state from
1688 * filtering, or because some other part of the kernel write an audit
1689 * message), then write out the syscall information.  In call cases,
1690 * free the names stored from getname().
1691 */
1692void audit_syscall_exit(int valid, long return_code)
1693{
1694        struct task_struct *tsk = current;
1695        struct audit_context *context;
1696
1697        context = audit_get_context(tsk, valid, return_code);
1698
1699        if (likely(!context))
1700                return;
1701
1702        if (context->in_syscall && context->current_state == AUDIT_RECORD_CONTEXT)
1703                audit_log_exit(context, tsk);
1704
1705        context->in_syscall = 0;
1706        context->prio = context->state == AUDIT_RECORD_CONTEXT ? ~0ULL : 0;
1707
1708        if (!list_empty(&context->killed_trees))
1709                audit_kill_trees(&context->killed_trees);
1710
1711        if (context->previous) {
1712                struct audit_context *new_context = context->previous;
1713                context->previous  = NULL;
1714                audit_free_context(context);
1715                tsk->audit_context = new_context;
1716        } else {
1717                audit_free_names(context);
1718                unroll_tree_refs(context, NULL, 0);
1719                audit_free_aux(context);
1720                context->aux = NULL;
1721                context->aux_pids = NULL;
1722                context->target_pid = 0;
1723                context->target_sid = 0;
1724                context->sockaddr_len = 0;
1725                context->type = 0;
1726                context->fds[0] = -1;
1727                if (context->state != AUDIT_RECORD_CONTEXT) {
1728                        kfree(context->filterkey);
1729                        context->filterkey = NULL;
1730                }
1731                tsk->audit_context = context;
1732        }
1733}
1734
1735static inline void handle_one(const struct inode *inode)
1736{
1737#ifdef CONFIG_AUDIT_TREE
1738        struct audit_context *context;
1739        struct audit_tree_refs *p;
1740        struct audit_chunk *chunk;
1741        int count;
1742        if (likely(hlist_empty(&inode->i_fsnotify_marks)))
1743                return;
1744        context = current->audit_context;
1745        p = context->trees;
1746        count = context->tree_count;
1747        rcu_read_lock();
1748        chunk = audit_tree_lookup(inode);
1749        rcu_read_unlock();
1750        if (!chunk)
1751                return;
1752        if (likely(put_tree_ref(context, chunk)))
1753                return;
1754        if (unlikely(!grow_tree_refs(context))) {
1755                printk(KERN_WARNING "out of memory, audit has lost a tree reference\n");
1756                audit_set_auditable(context);
1757                audit_put_chunk(chunk);
1758                unroll_tree_refs(context, p, count);
1759                return;
1760        }
1761        put_tree_ref(context, chunk);
1762#endif
1763}
1764
1765static void handle_path(const struct dentry *dentry)
1766{
1767#ifdef CONFIG_AUDIT_TREE
1768        struct audit_context *context;
1769        struct audit_tree_refs *p;
1770        const struct dentry *d, *parent;
1771        struct audit_chunk *drop;
1772        unsigned long seq;
1773        int count;
1774
1775        context = current->audit_context;
1776        p = context->trees;
1777        count = context->tree_count;
1778retry:
1779        drop = NULL;
1780        d = dentry;
1781        rcu_read_lock();
1782        seq = read_seqbegin(&rename_lock);
1783        for(;;) {
1784                struct inode *inode = d->d_inode;
1785                if (inode && unlikely(!hlist_empty(&inode->i_fsnotify_marks))) {
1786                        struct audit_chunk *chunk;
1787                        chunk = audit_tree_lookup(inode);
1788                        if (chunk) {
1789                                if (unlikely(!put_tree_ref(context, chunk))) {
1790                                        drop = chunk;
1791                                        break;
1792                                }
1793                        }
1794                }
1795                parent = d->d_parent;
1796                if (parent == d)
1797                        break;
1798                d = parent;
1799        }
1800        if (unlikely(read_seqretry(&rename_lock, seq) || drop)) {  /* in this order */
1801                rcu_read_unlock();
1802                if (!drop) {
1803                        /* just a race with rename */
1804                        unroll_tree_refs(context, p, count);
1805                        goto retry;
1806                }
1807                audit_put_chunk(drop);
1808                if (grow_tree_refs(context)) {
1809                        /* OK, got more space */
1810                        unroll_tree_refs(context, p, count);
1811                        goto retry;
1812                }
1813                /* too bad */
1814                printk(KERN_WARNING
1815                        "out of memory, audit has lost a tree reference\n");
1816                unroll_tree_refs(context, p, count);
1817                audit_set_auditable(context);
1818                return;
1819        }
1820        rcu_read_unlock();
1821#endif
1822}
1823
1824/**
1825 * audit_getname - add a name to the list
1826 * @name: name to add
1827 *
1828 * Add a name to the list of audit names for this context.
1829 * Called from fs/namei.c:getname().
1830 */
1831void __audit_getname(const char *name)
1832{
1833        struct audit_context *context = current->audit_context;
1834
1835        if (IS_ERR(name) || !name)
1836                return;
1837
1838        if (!context->in_syscall) {
1839#if AUDIT_DEBUG == 2
1840                printk(KERN_ERR "%s:%d(:%d): ignoring getname(%p)\n",
1841                       __FILE__, __LINE__, context->serial, name);
1842                dump_stack();
1843#endif
1844                return;
1845        }
1846        BUG_ON(context->name_count >= AUDIT_NAMES);
1847        context->names[context->name_count].name = name;
1848        context->names[context->name_count].name_len = AUDIT_NAME_FULL;
1849        context->names[context->name_count].name_put = 1;
1850        context->names[context->name_count].ino  = (unsigned long)-1;
1851        context->names[context->name_count].osid = 0;
1852        ++context->name_count;
1853        if (!context->pwd.dentry)
1854                get_fs_pwd(current->fs, &context->pwd);
1855}
1856
1857/* audit_putname - intercept a putname request
1858 * @name: name to intercept and delay for putname
1859 *
1860 * If we have stored the name from getname in the audit context,
1861 * then we delay the putname until syscall exit.
1862 * Called from include/linux/fs.h:putname().
1863 */
1864void audit_putname(const char *name)
1865{
1866        struct audit_context *context = current->audit_context;
1867
1868        BUG_ON(!context);
1869        if (!context->in_syscall) {
1870#if AUDIT_DEBUG == 2
1871                printk(KERN_ERR "%s:%d(:%d): __putname(%p)\n",
1872                       __FILE__, __LINE__, context->serial, name);
1873                if (context->name_count) {
1874                        int i;
1875                        for (i = 0; i < context->name_count; i++)
1876                                printk(KERN_ERR "name[%d] = %p = %s\n", i,
1877                                       context->names[i].name,
1878                                       context->names[i].name ?: "(null)");
1879                }
1880#endif
1881                __putname(name);
1882        }
1883#if AUDIT_DEBUG
1884        else {
1885                ++context->put_count;
1886                if (context->put_count > context->name_count) {
1887                        printk(KERN_ERR "%s:%d(:%d): major=%d"
1888                               " in_syscall=%d putname(%p) name_count=%d"
1889                               " put_count=%d\n",
1890                               __FILE__, __LINE__,
1891                               context->serial, context->major,
1892                               context->in_syscall, name, context->name_count,
1893                               context->put_count);
1894                        dump_stack();
1895                }
1896        }
1897#endif
1898}
1899
1900static int audit_inc_name_count(struct audit_context *context,
1901                                const struct inode *inode)
1902{
1903        if (context->name_count >= AUDIT_NAMES) {
1904                if (inode)
1905                        printk(KERN_DEBUG "audit: name_count maxed, losing inode data: "
1906                               "dev=%02x:%02x, inode=%lu\n",
1907                               MAJOR(inode->i_sb->s_dev),
1908                               MINOR(inode->i_sb->s_dev),
1909                               inode->i_ino);
1910
1911                else
1912                        printk(KERN_DEBUG "name_count maxed, losing inode data\n");
1913                return 1;
1914        }
1915        context->name_count++;
1916#if AUDIT_DEBUG
1917        context->ino_count++;
1918#endif
1919        return 0;
1920}
1921
1922
1923static inline int audit_copy_fcaps(struct audit_names *name, const struct dentry *dentry)
1924{
1925        struct cpu_vfs_cap_data caps;
1926        int rc;
1927
1928        memset(&name->fcap.permitted, 0, sizeof(kernel_cap_t));
1929        memset(&name->fcap.inheritable, 0, sizeof(kernel_cap_t));
1930        name->fcap.fE = 0;
1931        name->fcap_ver = 0;
1932
1933        if (!dentry)
1934                return 0;
1935
1936        rc = get_vfs_caps_from_disk(dentry, &caps);
1937        if (rc)
1938                return rc;
1939
1940        name->fcap.permitted = caps.permitted;
1941        name->fcap.inheritable = caps.inheritable;
1942        name->fcap.fE = !!(caps.magic_etc & VFS_CAP_FLAGS_EFFECTIVE);
1943        name->fcap_ver = (caps.magic_etc & VFS_CAP_REVISION_MASK) >> VFS_CAP_REVISION_SHIFT;
1944
1945        return 0;
1946}
1947
1948
1949/* Copy inode data into an audit_names. */
1950static void audit_copy_inode(struct audit_names *name, const struct dentry *dentry,
1951                             const struct inode *inode)
1952{
1953        name->ino   = inode->i_ino;
1954        name->dev   = inode->i_sb->s_dev;
1955        name->mode  = inode->i_mode;
1956        name->uid   = inode->i_uid;
1957        name->gid   = inode->i_gid;
1958        name->rdev  = inode->i_rdev;
1959        security_inode_getsecid(inode, &name->osid);
1960        audit_copy_fcaps(name, dentry);
1961}
1962
1963/**
1964 * audit_inode - store the inode and device from a lookup
1965 * @name: name being audited
1966 * @dentry: dentry being audited
1967 *
1968 * Called from fs/namei.c:path_lookup().
1969 */
1970void __audit_inode(const char *name, const struct dentry *dentry)
1971{
1972        int idx;
1973        struct audit_context *context = current->audit_context;
1974        const struct inode *inode = dentry->d_inode;
1975
1976        if (!context->in_syscall)
1977                return;
1978        if (context->name_count
1979            && context->names[context->name_count-1].name
1980            && context->names[context->name_count-1].name == name)
1981                idx = context->name_count - 1;
1982        else if (context->name_count > 1
1983                 && context->names[context->name_count-2].name
1984                 && context->names[context->name_count-2].name == name)
1985                idx = context->name_count - 2;
1986        else {
1987                /* FIXME: how much do we care about inodes that have no
1988                 * associated name? */
1989                if (audit_inc_name_count(context, inode))
1990                        return;
1991                idx = context->name_count - 1;
1992                context->names[idx].name = NULL;
1993        }
1994        handle_path(dentry);
1995        audit_copy_inode(&context->names[idx], dentry, inode);
1996}
1997
1998/**
1999 * audit_inode_child - collect inode info for created/removed objects
2000 * @dentry: dentry being audited
2001 * @parent: inode of dentry parent
2002 *
2003 * For syscalls that create or remove filesystem objects, audit_inode
2004 * can only collect information for the filesystem object's parent.
2005 * This call updates the audit context with the child's information.
2006 * Syscalls that create a new filesystem object must be hooked after
2007 * the object is created.  Syscalls that remove a filesystem object
2008 * must be hooked prior, in order to capture the target inode during
2009 * unsuccessful attempts.
2010 */
2011void __audit_inode_child(const struct dentry *dentry,
2012                         const struct inode *parent)
2013{
2014        int idx;
2015        struct audit_context *context = current->audit_context;
2016        const char *found_parent = NULL, *found_child = NULL;
2017        const struct inode *inode = dentry->d_inode;
2018        const char *dname = dentry->d_name.name;
2019        int dirlen = 0;
2020
2021        if (!context->in_syscall)
2022                return;
2023
2024        if (inode)
2025                handle_one(inode);
2026
2027        /* parent is more likely, look for it first */
2028        for (idx = 0; idx < context->name_count; idx++) {
2029                struct audit_names *n = &context->names[idx];
2030
2031                if (!n->name)
2032                        continue;
2033
2034                if (n->ino == parent->i_ino &&
2035                    !audit_compare_dname_path(dname, n->name, &dirlen)) {
2036                        n->name_len = dirlen; /* update parent data in place */
2037                        found_parent = n->name;
2038                        goto add_names;
2039                }
2040        }
2041
2042        /* no matching parent, look for matching child */
2043        for (idx = 0; idx < context->name_count; idx++) {
2044                struct audit_names *n = &context->names[idx];
2045
2046                if (!n->name)
2047                        continue;
2048
2049                /* strcmp() is the more likely scenario */
2050                if (!strcmp(dname, n->name) ||
2051                     !audit_compare_dname_path(dname, n->name, &dirlen)) {
2052                        if (inode)
2053                                audit_copy_inode(n, NULL, inode);
2054                        else
2055                                n->ino = (unsigned long)-1;
2056                        found_child = n->name;
2057                        goto add_names;
2058                }
2059        }
2060
2061add_names:
2062        if (!found_parent) {
2063                if (audit_inc_name_count(context, parent))
2064                        return;
2065                idx = context->name_count - 1;
2066                context->names[idx].name = NULL;
2067                audit_copy_inode(&context->names[idx], NULL, parent);
2068        }
2069
2070        if (!found_child) {
2071                if (audit_inc_name_count(context, inode))
2072                        return;
2073                idx = context->name_count - 1;
2074
2075                /* Re-use the name belonging to the slot for a matching parent
2076                 * directory. All names for this context are relinquished in
2077                 * audit_free_names() */
2078                if (found_parent) {
2079                        context->names[idx].name = found_parent;
2080                        context->names[idx].name_len = AUDIT_NAME_FULL;
2081                        /* don't call __putname() */
2082                        context->names[idx].name_put = 0;
2083                } else {
2084                        context->names[idx].name = NULL;
2085                }
2086
2087                if (inode)
2088                        audit_copy_inode(&context->names[idx], NULL, inode);
2089                else
2090                        context->names[idx].ino = (unsigned long)-1;
2091        }
2092}
2093EXPORT_SYMBOL_GPL(__audit_inode_child);
2094
2095/**
2096 * auditsc_get_stamp - get local copies of audit_context values
2097 * @ctx: audit_context for the task
2098 * @t: timespec to store time recorded in the audit_context
2099 * @serial: serial value that is recorded in the audit_context
2100 *
2101 * Also sets the context as auditable.
2102 */
2103int auditsc_get_stamp(struct audit_context *ctx,
2104                       struct timespec *t, unsigned int *serial)
2105{
2106        if (!ctx->in_syscall)
2107                return 0;
2108        if (!ctx->serial)
2109                ctx->serial = audit_serial();
2110        t->tv_sec  = ctx->ctime.tv_sec;
2111        t->tv_nsec = ctx->ctime.tv_nsec;
2112        *serial    = ctx->serial;
2113        if (!ctx->prio) {
2114                ctx->prio = 1;
2115                ctx->current_state = AUDIT_RECORD_CONTEXT;
2116        }
2117        return 1;
2118}
2119
2120/* global counter which is incremented every time something logs in */
2121static atomic_t session_id = ATOMIC_INIT(0);
2122
2123/**
2124 * audit_set_loginuid - set a task's audit_context loginuid
2125 * @task: task whose audit context is being modified
2126 * @loginuid: loginuid value
2127 *
2128 * Returns 0.
2129 *
2130 * Called (set) from fs/proc/base.c::proc_loginuid_write().
2131 */
2132int audit_set_loginuid(struct task_struct *task, uid_t loginuid)
2133{
2134        unsigned int sessionid = atomic_inc_return(&session_id);
2135        struct audit_context *context = task->audit_context;
2136
2137        if (context && context->in_syscall) {
2138                struct audit_buffer *ab;
2139
2140                ab = audit_log_start(NULL, GFP_KERNEL, AUDIT_LOGIN);
2141                if (ab) {
2142                        audit_log_format(ab, "login pid=%d uid=%u "
2143                                "old auid=%u new auid=%u"
2144                                " old ses=%u new ses=%u",
2145                                task->pid, task_uid(task),
2146                                task->loginuid, loginuid,
2147                                task->sessionid, sessionid);
2148                        audit_log_end(ab);
2149                }
2150        }
2151        task->sessionid = sessionid;
2152        task->loginuid = loginuid;
2153        return 0;
2154}
2155
2156/**
2157 * __audit_mq_open - record audit data for a POSIX MQ open
2158 * @oflag: open flag
2159 * @mode: mode bits
2160 * @attr: queue attributes
2161 *
2162 */
2163void __audit_mq_open(int oflag, mode_t mode, struct mq_attr *attr)
2164{
2165        struct audit_context *context = current->audit_context;
2166
2167        if (attr)
2168                memcpy(&context->mq_open.attr, attr, sizeof(struct mq_attr));
2169        else
2170                memset(&context->mq_open.attr, 0, sizeof(struct mq_attr));
2171
2172        context->mq_open.oflag = oflag;
2173        context->mq_open.mode = mode;
2174
2175        context->type = AUDIT_MQ_OPEN;
2176}
2177
2178/**
2179 * __audit_mq_sendrecv - record audit data for a POSIX MQ timed send/receive
2180 * @mqdes: MQ descriptor
2181 * @msg_len: Message length
2182 * @msg_prio: Message priority
2183 * @abs_timeout: Message timeout in absolute time
2184 *
2185 */
2186void __audit_mq_sendrecv(mqd_t mqdes, size_t msg_len, unsigned int msg_prio,
2187                        const struct timespec *abs_timeout)
2188{
2189        struct audit_context *context = current->audit_context;
2190        struct timespec *p = &context->mq_sendrecv.abs_timeout;
2191
2192        if (abs_timeout)
2193                memcpy(p, abs_timeout, sizeof(struct timespec));
2194        else
2195                memset(p, 0, sizeof(struct timespec));
2196
2197        context->mq_sendrecv.mqdes = mqdes;
2198        context->mq_sendrecv.msg_len = msg_len;
2199        context->mq_sendrecv.msg_prio = msg_prio;
2200
2201        context->type = AUDIT_MQ_SENDRECV;
2202}
2203
2204/**
2205 * __audit_mq_notify - record audit data for a POSIX MQ notify
2206 * @mqdes: MQ descriptor
2207 * @notification: Notification event
2208 *
2209 */
2210
2211void __audit_mq_notify(mqd_t mqdes, const struct sigevent *notification)
2212{
2213        struct audit_context *context = current->audit_context;
2214
2215        if (notification)
2216                context->mq_notify.sigev_signo = notification->sigev_signo;
2217        else
2218                context->mq_notify.sigev_signo = 0;
2219
2220        context->mq_notify.mqdes = mqdes;
2221        context->type = AUDIT_MQ_NOTIFY;
2222}
2223
2224/**
2225 * __audit_mq_getsetattr - record audit data for a POSIX MQ get/set attribute
2226 * @mqdes: MQ descriptor
2227 * @mqstat: MQ flags
2228 *
2229 */
2230void __audit_mq_getsetattr(mqd_t mqdes, struct mq_attr *mqstat)
2231{
2232        struct audit_context *context = current->audit_context;
2233        context->mq_getsetattr.mqdes = mqdes;
2234        context->mq_getsetattr.mqstat = *mqstat;
2235        context->type = AUDIT_MQ_GETSETATTR;
2236}
2237
2238/**
2239 * audit_ipc_obj - record audit data for ipc object
2240 * @ipcp: ipc permissions
2241 *
2242 */
2243void __audit_ipc_obj(struct kern_ipc_perm *ipcp)
2244{
2245        struct audit_context *context = current->audit_context;
2246        context->ipc.uid = ipcp->uid;
2247        context->ipc.gid = ipcp->gid;
2248        context->ipc.mode = ipcp->mode;
2249        context->ipc.has_perm = 0;
2250        security_ipc_getsecid(ipcp, &context->ipc.osid);
2251        context->type = AUDIT_IPC;
2252}
2253
2254/**
2255 * audit_ipc_set_perm - record audit data for new ipc permissions
2256 * @qbytes: msgq bytes
2257 * @uid: msgq user id
2258 * @gid: msgq group id
2259 * @mode: msgq mode (permissions)
2260 *
2261 * Called only after audit_ipc_obj().
2262 */
2263void __audit_ipc_set_perm(unsigned long qbytes, uid_t uid, gid_t gid, mode_t mode)
2264{
2265        struct audit_context *context = current->audit_context;
2266
2267        context->ipc.qbytes = qbytes;
2268        context->ipc.perm_uid = uid;
2269        context->ipc.perm_gid = gid;
2270        context->ipc.perm_mode = mode;
2271        context->ipc.has_perm = 1;
2272}
2273
2274int audit_bprm(struct linux_binprm *bprm)
2275{
2276        struct audit_aux_data_execve *ax;
2277        struct audit_context *context = current->audit_context;
2278
2279        if (likely(!audit_enabled || !context || context->dummy))
2280                return 0;
2281
2282        ax = kmalloc(sizeof(*ax), GFP_KERNEL);
2283        if (!ax)
2284                return -ENOMEM;
2285
2286        ax->argc = bprm->argc;
2287        ax->envc = bprm->envc;
2288        ax->mm = bprm->mm;
2289        ax->d.type = AUDIT_EXECVE;
2290        ax->d.next = context->aux;
2291        context->aux = (void *)ax;
2292        return 0;
2293}
2294
2295
2296/**
2297 * audit_socketcall - record audit data for sys_socketcall
2298 * @nargs: number of args
2299 * @args: args array
2300 *
2301 */
2302void audit_socketcall(int nargs, unsigned long *args)
2303{
2304        struct audit_context *context = current->audit_context;
2305
2306        if (likely(!context || context->dummy))
2307                return;
2308
2309        context->type = AUDIT_SOCKETCALL;
2310        context->socketcall.nargs = nargs;
2311        memcpy(context->socketcall.args, args, nargs * sizeof(unsigned long));
2312}
2313
2314/**
2315 * __audit_fd_pair - record audit data for pipe and socketpair
2316 * @fd1: the first file descriptor
2317 * @fd2: the second file descriptor
2318 *
2319 */
2320void __audit_fd_pair(int fd1, int fd2)
2321{
2322        struct audit_context *context = current->audit_context;
2323        context->fds[0] = fd1;
2324        context->fds[1] = fd2;
2325}
2326
2327/**
2328 * audit_sockaddr - record audit data for sys_bind, sys_connect, sys_sendto
2329 * @len: data length in user space
2330 * @a: data address in kernel space
2331 *
2332 * Returns 0 for success or NULL context or < 0 on error.
2333 */
2334int audit_sockaddr(int len, void *a)
2335{
2336        struct audit_context *context = current->audit_context;
2337
2338        if (likely(!context || context->dummy))
2339                return 0;
2340
2341        if (!context->sockaddr) {
2342                void *p = kmalloc(sizeof(struct sockaddr_storage), GFP_KERNEL);
2343                if (!p)
2344                        return -ENOMEM;
2345                context->sockaddr = p;
2346        }
2347
2348        context->sockaddr_len = len;
2349        memcpy(context->sockaddr, a, len);
2350        return 0;
2351}
2352
2353void __audit_ptrace(struct task_struct *t)
2354{
2355        struct audit_context *context = current->audit_context;
2356
2357        context->target_pid = t->pid;
2358        context->target_auid = audit_get_loginuid(t);
2359        context->target_uid = task_uid(t);
2360        context->target_sessionid = audit_get_sessionid(t);
2361        security_task_getsecid(t, &context->target_sid);
2362        memcpy(context->target_comm, t->comm, TASK_COMM_LEN);
2363}
2364
2365/**
2366 * audit_signal_info - record signal info for shutting down audit subsystem
2367 * @sig: signal value
2368 * @t: task being signaled
2369 *
2370 * If the audit subsystem is being terminated, record the task (pid)
2371 * and uid that is doing that.
2372 */
2373int __audit_signal_info(int sig, struct task_struct *t)
2374{
2375        struct audit_aux_data_pids *axp;
2376        struct task_struct *tsk = current;
2377        struct audit_context *ctx = tsk->audit_context;
2378        uid_t uid = current_uid(), t_uid = task_uid(t);
2379
2380        if (audit_pid && t->tgid == audit_pid) {
2381                if (sig == SIGTERM || sig == SIGHUP || sig == SIGUSR1 || sig == SIGUSR2) {
2382                        audit_sig_pid = tsk->pid;
2383                        if (tsk->loginuid != -1)
2384                                audit_sig_uid = tsk->loginuid;
2385                        else
2386                                audit_sig_uid = uid;
2387                        security_task_getsecid(tsk, &audit_sig_sid);
2388                }
2389                if (!audit_signals || audit_dummy_context())
2390                        return 0;
2391        }
2392
2393        /* optimize the common case by putting first signal recipient directly
2394         * in audit_context */
2395        if (!ctx->target_pid) {
2396                ctx->target_pid = t->tgid;
2397                ctx->target_auid = audit_get_loginuid(t);
2398                ctx->target_uid = t_uid;
2399                ctx->target_sessionid = audit_get_sessionid(t);
2400                security_task_getsecid(t, &ctx->target_sid);
2401                memcpy(ctx->target_comm, t->comm, TASK_COMM_LEN);
2402                return 0;
2403        }
2404
2405        axp = (void *)ctx->aux_pids;
2406        if (!axp || axp->pid_count == AUDIT_AUX_PIDS) {
2407                axp = kzalloc(sizeof(*axp), GFP_ATOMIC);
2408                if (!axp)
2409                        return -ENOMEM;
2410
2411                axp->d.type = AUDIT_OBJ_PID;
2412                axp->d.next = ctx->aux_pids;
2413                ctx->aux_pids = (void *)axp;
2414        }
2415        BUG_ON(axp->pid_count >= AUDIT_AUX_PIDS);
2416
2417        axp->target_pid[axp->pid_count] = t->tgid;
2418        axp->target_auid[axp->pid_count] = audit_get_loginuid(t);
2419        axp->target_uid[axp->pid_count] = t_uid;
2420        axp->target_sessionid[axp->pid_count] = audit_get_sessionid(t);
2421        security_task_getsecid(t, &axp->target_sid[axp->pid_count]);
2422        memcpy(axp->target_comm[axp->pid_count], t->comm, TASK_COMM_LEN);
2423        axp->pid_count++;
2424
2425        return 0;
2426}
2427
2428/**
2429 * __audit_log_bprm_fcaps - store information about a loading bprm and relevant fcaps
2430 * @bprm: pointer to the bprm being processed
2431 * @new: the proposed new credentials
2432 * @old: the old credentials
2433 *
2434 * Simply check if the proc already has the caps given by the file and if not
2435 * store the priv escalation info for later auditing at the end of the syscall
2436 *
2437 * -Eric
2438 */
2439int __audit_log_bprm_fcaps(struct linux_binprm *bprm,
2440                           const struct cred *new, const struct cred *old)
2441{
2442        struct audit_aux_data_bprm_fcaps *ax;
2443        struct audit_context *context = current->audit_context;
2444        struct cpu_vfs_cap_data vcaps;
2445        struct dentry *dentry;
2446
2447        ax = kmalloc(sizeof(*ax), GFP_KERNEL);
2448        if (!ax)
2449                return -ENOMEM;
2450
2451        ax->d.type = AUDIT_BPRM_FCAPS;
2452        ax->d.next = context->aux;
2453        context->aux = (void *)ax;
2454
2455        dentry = dget(bprm->file->f_dentry);
2456        get_vfs_caps_from_disk(dentry, &vcaps);
2457        dput(dentry);
2458
2459        ax->fcap.permitted = vcaps.permitted;
2460        ax->fcap.inheritable = vcaps.inheritable;
2461        ax->fcap.fE = !!(vcaps.magic_etc & VFS_CAP_FLAGS_EFFECTIVE);
2462        ax->fcap_ver = (vcaps.magic_etc & VFS_CAP_REVISION_MASK) >> VFS_CAP_REVISION_SHIFT;
2463
2464        ax->old_pcap.permitted   = old->cap_permitted;
2465        ax->old_pcap.inheritable = old->cap_inheritable;
2466        ax->old_pcap.effective   = old->cap_effective;
2467
2468        ax->new_pcap.permitted   = new->cap_permitted;
2469        ax->new_pcap.inheritable = new->cap_inheritable;
2470        ax->new_pcap.effective   = new->cap_effective;
2471        return 0;
2472}
2473
2474/**
2475 * __audit_log_capset - store information about the arguments to the capset syscall
2476 * @pid: target pid of the capset call
2477 * @new: the new credentials
2478 * @old: the old (current) credentials
2479 *
2480 * Record the aguments userspace sent to sys_capset for later printing by the
2481 * audit system if applicable
2482 */
2483void __audit_log_capset(pid_t pid,
2484                       const struct cred *new, const struct cred *old)
2485{
2486        struct audit_context *context = current->audit_context;
2487        context->capset.pid = pid;
2488        context->capset.cap.effective   = new->cap_effective;
2489        context->capset.cap.inheritable = new->cap_effective;
2490        context->capset.cap.permitted   = new->cap_permitted;
2491        context->type = AUDIT_CAPSET;
2492}
2493
2494void __audit_mmap_fd(int fd, int flags)
2495{
2496        struct audit_context *context = current->audit_context;
2497        context->mmap.fd = fd;
2498        context->mmap.flags = flags;
2499        context->type = AUDIT_MMAP;
2500}
2501
2502/**
2503 * audit_core_dumps - record information about processes that end abnormally
2504 * @signr: signal value
2505 *
2506 * If a process ends with a core dump, something fishy is going on and we
2507 * should record the event for investigation.
2508 */
2509void audit_core_dumps(long signr)
2510{
2511        struct audit_buffer *ab;
2512        u32 sid;
2513        uid_t auid = audit_get_loginuid(current), uid;
2514        gid_t gid;
2515        unsigned int sessionid = audit_get_sessionid(current);
2516
2517        if (!audit_enabled)
2518                return;
2519
2520        if (signr == SIGQUIT)   /* don't care for those */
2521                return;
2522
2523        ab = audit_log_start(NULL, GFP_KERNEL, AUDIT_ANOM_ABEND);
2524        current_uid_gid(&uid, &gid);
2525        audit_log_format(ab, "auid=%u uid=%u gid=%u ses=%u",
2526                         auid, uid, gid, sessionid);
2527        security_task_getsecid(current, &sid);
2528        if (sid) {
2529                char *ctx = NULL;
2530                u32 len;
2531
2532                if (security_secid_to_secctx(sid, &ctx, &len))
2533                        audit_log_format(ab, " ssid=%u", sid);
2534                else {
2535                        audit_log_format(ab, " subj=%s", ctx);
2536                        security_release_secctx(ctx, len);
2537                }
2538        }
2539        audit_log_format(ab, " pid=%d comm=", current->pid);
2540        audit_log_untrustedstring(ab, current->comm);
2541        audit_log_format(ab, " sig=%ld", signr);
2542        audit_log_end(ab);
2543}
2544
2545struct list_head *audit_killed_trees(void)
2546{
2547        struct audit_context *ctx = current->audit_context;
2548        if (likely(!ctx || !ctx->in_syscall))
2549                return NULL;
2550        return &ctx->killed_trees;
2551}
2552
lxr.linux.no kindly hosted by Redpill Linpro AS, provider of Linux consulting and operations services since 1995.