kernel/sysctl.c

   1 /*
   2  * sysctl.c: General linux system control interface
   3  *
   4  * Begun 24 March 1995, Stephen Tweedie
   5  * Added /proc support, Dec 1995
   6  * Added bdflush entry and intvec min/max checking, 2/23/96, Tom Dyas.
   7  * Added hooks for /proc/sys/net (minor, minor patch), 96/4/1, Mike Shaver.
   8  * Added kernel/java-{interpreter,appletviewer}, 96/5/10, Mike Shaver.
   9  * Dynamic registration fixes, Stephen Tweedie.
  10  * Added kswapd-interval, ctrl-alt-del, printk stuff, 1/8/97, Chris Horn.
  11  * Made sysctl support optional via CONFIG_SYSCTL, 1/10/97, Chris
  12  *  Horn.
  13  * Added proc_doulongvec_ms_jiffies_minmax, 09/08/99, Carlos H. Bauer.
  14  * Added proc_doulongvec_minmax, 09/08/99, Carlos H. Bauer.
  15  * Changed linked lists to use list.h instead of lists.h, 02/24/00, Bill
  16  *  Wendling.
  17  * The list_for_each() macro wasn't appropriate for the sysctl loop.
  18  *  Removed it and replaced it with older style, 03/23/00, Bill Wendling
  19  */
  20
  21 #include <linux/module.h>
  22 #include <linux/mm.h>
  23 #include <linux/swap.h>
  24 #include <linux/slab.h>
  25 #include <linux/sysctl.h>
  26 #include <linux/signal.h>
  27 #include <linux/proc_fs.h>
  28 #include <linux/security.h>
  29 #include <linux/ctype.h>
  30 #include <linux/kmemcheck.h>
  31 #include <linux/fs.h>
  32 #include <linux/init.h>
  33 #include <linux/kernel.h>
  34 #include <linux/kobject.h>
  35 #include <linux/net.h>
  36 #include <linux/sysrq.h>
  37 #include <linux/highuid.h>
  38 #include <linux/writeback.h>
  39 #include <linux/ratelimit.h>
  40 #include <linux/hugetlb.h>
  41 #include <linux/initrd.h>
  42 #include <linux/key.h>
  43 #include <linux/times.h>
  44 #include <linux/limits.h>
  45 #include <linux/dcache.h>
  46 #include <linux/syscalls.h>
  47 #include <linux/vmstat.h>
  48 #include <linux/nfs_fs.h>
  49 #include <linux/acpi.h>
  50 #include <linux/reboot.h>
  51 #include <linux/ftrace.h>
  52 #include <linux/slow-work.h>
  53 #include <linux/perf_event.h>
  54 #include <linux/kprobes.h>
  55
  56 #include <asm/uaccess.h>
  57 #include <asm/processor.h>
  58
  59 #ifdef CONFIG_X86
  60 #include <asm/nmi.h>
  61 #include <asm/stacktrace.h>
  62 #include <asm/io.h>
  63 #endif
  64 #ifdef CONFIG_BSD_PROCESS_ACCT
  65 #include <linux/acct.h>
  66 #endif
  67 #ifdef CONFIG_RT_MUTEXES
  68 #include <linux/rtmutex.h>
  69 #endif
  70 #if defined(CONFIG_PROVE_LOCKING) || defined(CONFIG_LOCK_STAT)
  71 #include <linux/lockdep.h>
  72 #endif
  73 #ifdef CONFIG_CHR_DEV_SG
  74 #include <scsi/sg.h>
  75 #endif
  76
  77
  78 #if defined(CONFIG_SYSCTL)
  79
  80 /* External variables not in a header file. */
  81 extern int sysctl_overcommit_memory;
  82 extern int sysctl_overcommit_ratio;
  83 extern int sysctl_panic_on_oom;
  84 extern int sysctl_oom_kill_allocating_task;
  85 extern int sysctl_oom_dump_tasks;
  86 extern int max_threads;
  87 extern int core_uses_pid;
  88 extern int suid_dumpable;
  89 extern char core_pattern[];
  90 extern unsigned int core_pipe_limit;
  91 extern int pid_max;
  92 extern int min_free_kbytes;
  93 extern int pid_max_min, pid_max_max;
  94 extern int sysctl_drop_caches;
  95 extern int percpu_pagelist_fraction;
  96 extern int compat_log;
  97 extern int latencytop_enabled;
  98 extern int sysctl_nr_open_min, sysctl_nr_open_max;
  99 #ifndef CONFIG_MMU
 100 extern int sysctl_nr_trim_pages;
 101 #endif
 102 #ifdef CONFIG_BLOCK
 103 extern int blk_iopoll_enabled;
 104 #endif
 105
 106 /* Constants used for minimum and  maximum */
 107 #ifdef CONFIG_DETECT_SOFTLOCKUP
 108 static int sixty = 60;
 109 static int neg_one = -1;
 110 #endif
 111
 112 static int zero;
 113 static int __maybe_unused one = 1;
 114 static int __maybe_unused two = 2;
 115 static unsigned long one_ul = 1;
 116 static int one_hundred = 100;
 117 #ifdef CONFIG_PRINTK
 118 static int ten_thousand = 10000;
 119 #endif
 120
 121 /* this is needed for the proc_doulongvec_minmax of vm_dirty_bytes */
 122 static unsigned long dirty_bytes_min = 2 * PAGE_SIZE;
 123
 124 /* this is needed for the proc_dointvec_minmax for [fs_]overflow UID and GID */
 125 static int maxolduid = 65535;
 126 static int minolduid;
 127 static int min_percpu_pagelist_fract = 8;
 128
 129 static int ngroups_max = NGROUPS_MAX;
 130
 131 #ifdef CONFIG_SPARC
 132 #include <asm/system.h>
 133 #endif
 134
 135 #ifdef CONFIG_SPARC64
 136 extern int sysctl_tsb_ratio;
 137 #endif
 138
 139 #ifdef __hppa__
 140 extern int pwrsw_enabled;
 141 extern int unaligned_enabled;
 142 #endif
 143
 144 #ifdef CONFIG_S390
 145 #ifdef CONFIG_MATHEMU
 146 extern int sysctl_ieee_emulation_warnings;
 147 #endif
 148 extern int sysctl_userprocess_debug;
 149 extern int spin_retry;
 150 #endif
 151
 152 #ifdef CONFIG_IA64
 153 extern int no_unaligned_warning;
 154 extern int unaligned_dump_stack;
 155 #endif
 156
 157 extern struct ratelimit_state printk_ratelimit_state;
 158
 159 #ifdef CONFIG_PROC_SYSCTL
 160 static int proc_do_cad_pid(struct ctl_table *table, int write,
 161                   void __user *buffer, size_t *lenp, loff_t *ppos);
 162 static int proc_taint(struct ctl_table *table, int write,
 163                                void __user *buffer, size_t *lenp, loff_t *ppos);
 164 #endif
 165
 166 #ifdef CONFIG_MAGIC_SYSRQ
 167 static int __sysrq_enabled; /* Note: sysrq code ises it's own private copy */
 168
 169 static int sysrq_sysctl_handler(ctl_table *table, int write,
 170                                 void __user *buffer, size_t *lenp,
 171                                 loff_t *ppos)
 172 {
 173         int error;
 174
 175         error = proc_dointvec(table, write, buffer, lenp, ppos);
 176         if (error)
 177                 return error;
 178
 179         if (write)
 180                 sysrq_toggle_support(__sysrq_enabled);
 181
 182         return 0;
 183 }
 184
 185 #endif
 186
 187 static struct ctl_table root_table[];
 188 static struct ctl_table_root sysctl_table_root;
 189 static struct ctl_table_header root_table_header = {
 190         .count = 1,
 191         .ctl_table = root_table,
 192         .ctl_entry = LIST_HEAD_INIT(sysctl_table_root.default_set.list),
 193         .root = &sysctl_table_root,
 194         .set = &sysctl_table_root.default_set,
 195 };
 196 static struct ctl_table_root sysctl_table_root = {
 197         .root_list = LIST_HEAD_INIT(sysctl_table_root.root_list),
 198         .default_set.list = LIST_HEAD_INIT(root_table_header.ctl_entry),
 199 };
 200
 201 static struct ctl_table kern_table[];
 202 static struct ctl_table vm_table[];
 203 static struct ctl_table fs_table[];
 204 static struct ctl_table debug_table[];
 205 static struct ctl_table dev_table[];
 206 extern struct ctl_table random_table[];
 207 #ifdef CONFIG_INOTIFY_USER
 208 extern struct ctl_table inotify_table[];
 209 #endif
 210 #ifdef CONFIG_EPOLL
 211 extern struct ctl_table epoll_table[];
 212 #endif
 213
 214 #ifdef HAVE_ARCH_PICK_MMAP_LAYOUT
 215 int sysctl_legacy_va_layout;
 216 #endif
 217
 218 /* The default sysctl tables: */
 219
 220 static struct ctl_table root_table[] = {
 221         {
 222                 .procname       = "kernel",
 223                 .mode           = 0555,
 224                 .child          = kern_table,
 225         },
 226         {
 227                 .procname       = "vm",
 228                 .mode           = 0555,
 229                 .child          = vm_table,
 230         },
 231         {
 232                 .procname       = "fs",
 233                 .mode           = 0555,
 234                 .child          = fs_table,
 235         },
 236         {
 237                 .procname       = "debug",
 238                 .mode           = 0555,
 239                 .child          = debug_table,
 240         },
 241         {
 242                 .procname       = "dev",
 243                 .mode           = 0555,
 244                 .child          = dev_table,
 245         },
 246 /*
 247  * NOTE: do not add new entries to this table unless you have read
 248  * Documentation/sysctl/ctl_unnumbered.txt
 249  */
 250         { }
 251 };
 252
 253 #ifdef CONFIG_SCHED_DEBUG
 254 static int min_sched_granularity_ns = 100000;           /* 100 usecs */
 255 static int max_sched_granularity_ns = NSEC_PER_SEC;     /* 1 second */
 256 static int min_wakeup_granularity_ns;                   /* 0 usecs */
 257 static int max_wakeup_granularity_ns = NSEC_PER_SEC;    /* 1 second */
 258 static int min_sched_tunable_scaling = SCHED_TUNABLESCALING_NONE;
 259 static int max_sched_tunable_scaling = SCHED_TUNABLESCALING_END-1;
 260 static int min_sched_shares_ratelimit = 100000; /* 100 usec */
 261 static int max_sched_shares_ratelimit = NSEC_PER_SEC; /* 1 second */
 262 #endif
 263
 264 static struct ctl_table kern_table[] = {
 265         {
 266                 .procname       = "sched_child_runs_first",
 267                 .data           = &sysctl_sched_child_runs_first,
 268                 .maxlen         = sizeof(unsigned int),
 269                 .mode           = 0644,
 270                 .proc_handler   = proc_dointvec,
 271         },
 272 #ifdef CONFIG_SCHED_DEBUG
 273         {
 274                 .procname       = "sched_min_granularity_ns",
 275                 .data           = &sysctl_sched_min_granularity,
 276                 .maxlen         = sizeof(unsigned int),
 277                 .mode           = 0644,
 278                 .proc_handler   = sched_proc_update_handler,
 279                 .extra1         = &min_sched_granularity_ns,
 280                 .extra2         = &max_sched_granularity_ns,
 281         },
 282         {
 283                 .procname       = "sched_latency_ns",
 284                 .data           = &sysctl_sched_latency,
 285                 .maxlen         = sizeof(unsigned int),
 286                 .mode           = 0644,
 287                 .proc_handler   = sched_proc_update_handler,
 288                 .extra1         = &min_sched_granularity_ns,
 289                 .extra2         = &max_sched_granularity_ns,
 290         },
 291         {
 292                 .procname       = "sched_wakeup_granularity_ns",
 293                 .data           = &sysctl_sched_wakeup_granularity,
 294                 .maxlen         = sizeof(unsigned int),
 295                 .mode           = 0644,
 296                 .proc_handler   = sched_proc_update_handler,
 297                 .extra1         = &min_wakeup_granularity_ns,
 298                 .extra2         = &max_wakeup_granularity_ns,
 299         },
 300         {
 301                 .procname       = "sched_shares_ratelimit",
 302                 .data           = &sysctl_sched_shares_ratelimit,
 303                 .maxlen         = sizeof(unsigned int),
 304                 .mode           = 0644,
 305                 .proc_handler   = sched_proc_update_handler,
 306                 .extra1         = &min_sched_shares_ratelimit,
 307                 .extra2         = &max_sched_shares_ratelimit,
 308         },
 309         {
 310                 .procname       = "sched_tunable_scaling",
 311                 .data           = &sysctl_sched_tunable_scaling,
 312                 .maxlen         = sizeof(enum sched_tunable_scaling),
 313                 .mode           = 0644,
 314                 .proc_handler   = sched_proc_update_handler,
 315                 .extra1         = &min_sched_tunable_scaling,
 316                 .extra2         = &max_sched_tunable_scaling,
 317         },
 318         {
 319                 .procname       = "sched_shares_thresh",
 320                 .data           = &sysctl_sched_shares_thresh,
 321                 .maxlen         = sizeof(unsigned int),
 322                 .mode           = 0644,
 323                 .proc_handler   = proc_dointvec_minmax,
 324                 .extra1         = &zero,
 325         },
 326         {
 327                 .procname       = "sched_migration_cost",
 328                 .data           = &sysctl_sched_migration_cost,
 329                 .maxlen         = sizeof(unsigned int),
 330                 .mode           = 0644,
 331                 .proc_handler   = proc_dointvec,
 332         },
 333         {
 334                 .procname       = "sched_nr_migrate",
 335                 .data           = &sysctl_sched_nr_migrate,
 336                 .maxlen         = sizeof(unsigned int),
 337                 .mode           = 0644,
 338                 .proc_handler   = proc_dointvec,
 339         },
 340         {
 341                 .procname       = "sched_time_avg",
 342                 .data           = &sysctl_sched_time_avg,
 343                 .maxlen         = sizeof(unsigned int),
 344                 .mode           = 0644,
 345                 .proc_handler   = proc_dointvec,
 346         },
 347         {
 348                 .procname       = "timer_migration",
 349                 .data           = &sysctl_timer_migration,
 350                 .maxlen         = sizeof(unsigned int),
 351                 .mode           = 0644,
 352                 .proc_handler   = proc_dointvec_minmax,
 353                 .extra1         = &zero,
 354                 .extra2         = &one,
 355         },
 356 #endif
 357         {
 358                 .procname       = "sched_rt_period_us",
 359                 .data           = &sysctl_sched_rt_period,
 360                 .maxlen         = sizeof(unsigned int),
 361                 .mode           = 0644,
 362                 .proc_handler   = sched_rt_handler,
 363         },
 364         {
 365                 .procname       = "sched_rt_runtime_us",
 366                 .data           = &sysctl_sched_rt_runtime,
 367                 .maxlen         = sizeof(int),
 368                 .mode           = 0644,
 369                 .proc_handler   = sched_rt_handler,
 370         },
 371         {
 372                 .procname       = "sched_compat_yield",
 373                 .data           = &sysctl_sched_compat_yield,
 374                 .maxlen         = sizeof(unsigned int),
 375                 .mode           = 0644,
 376                 .proc_handler   = proc_dointvec,
 377         },
 378 #ifdef CONFIG_PROVE_LOCKING
 379         {
 380                 .procname       = "prove_locking",
 381                 .data           = &prove_locking,
 382                 .maxlen         = sizeof(int),
 383                 .mode           = 0644,
 384                 .proc_handler   = proc_dointvec,
 385         },
 386 #endif
 387 #ifdef CONFIG_LOCK_STAT
 388         {
 389                 .procname       = "lock_stat",
 390                 .data           = &lock_stat,
 391                 .maxlen         = sizeof(int),
 392                 .mode           = 0644,
 393                 .proc_handler   = proc_dointvec,
 394         },
 395 #endif
 396         {
 397                 .procname       = "panic",
 398                 .data           = &panic_timeout,
 399                 .maxlen         = sizeof(int),
 400                 .mode           = 0644,
 401                 .proc_handler   = proc_dointvec,
 402         },
 403         {
 404                 .procname       = "core_uses_pid",
 405                 .data           = &core_uses_pid,
 406                 .maxlen         = sizeof(int),
 407                 .mode           = 0644,
 408                 .proc_handler   = proc_dointvec,
 409         },
 410         {
 411                 .procname       = "core_pattern",
 412                 .data           = core_pattern,
 413                 .maxlen         = CORENAME_MAX_SIZE,
 414                 .mode           = 0644,
 415                 .proc_handler   = proc_dostring,
 416         },
 417         {
 418                 .procname       = "core_pipe_limit",
 419                 .data           = &core_pipe_limit,
 420                 .maxlen         = sizeof(unsigned int),
 421                 .mode           = 0644,
 422                 .proc_handler   = proc_dointvec,
 423         },
 424 #ifdef CONFIG_PROC_SYSCTL
 425         {
 426                 .procname       = "tainted",
 427                 .maxlen         = sizeof(long),
 428                 .mode           = 0644,
 429                 .proc_handler   = proc_taint,
 430         },
 431 #endif
 432 #ifdef CONFIG_LATENCYTOP
 433         {
 434                 .procname       = "latencytop",
 435                 .data           = &latencytop_enabled,
 436                 .maxlen         = sizeof(int),
 437                 .mode           = 0644,
 438                 .proc_handler   = proc_dointvec,
 439         },
 440 #endif
 441 #ifdef CONFIG_BLK_DEV_INITRD
 442         {
 443                 .procname       = "real-root-dev",
 444                 .data           = &real_root_dev,
 445                 .maxlen         = sizeof(int),
 446                 .mode           = 0644,
 447                 .proc_handler   = proc_dointvec,
 448         },
 449 #endif
 450         {
 451                 .procname       = "print-fatal-signals",
 452                 .data           = &print_fatal_signals,
 453                 .maxlen         = sizeof(int),
 454                 .mode           = 0644,
 455                 .proc_handler   = proc_dointvec,
 456         },
 457 #ifdef CONFIG_SPARC
 458         {
 459                 .procname       = "reboot-cmd",
 460                 .data           = reboot_command,
 461                 .maxlen         = 256,
 462                 .mode           = 0644,
 463                 .proc_handler   = proc_dostring,
 464         },
 465         {
 466                 .procname       = "stop-a",
 467                 .data           = &stop_a_enabled,
 468                 .maxlen         = sizeof (int),
 469                 .mode           = 0644,
 470                 .proc_handler   = proc_dointvec,
 471         },
 472         {
 473                 .procname       = "scons-poweroff",
 474                 .data           = &scons_pwroff,
 475                 .maxlen         = sizeof (int),
 476                 .mode           = 0644,
 477                 .proc_handler   = proc_dointvec,
 478         },
 479 #endif
 480 #ifdef CONFIG_SPARC64
 481         {
 482                 .procname       = "tsb-ratio",
 483                 .data           = &sysctl_tsb_ratio,
 484                 .maxlen         = sizeof (int),
 485                 .mode           = 0644,
 486                 .proc_handler   = proc_dointvec,
 487         },
 488 #endif
 489 #ifdef __hppa__
 490         {
 491                 .procname       = "soft-power",
 492                 .data           = &pwrsw_enabled,
 493                 .maxlen         = sizeof (int),
 494                 .mode           = 0644,
 495                 .proc_handler   = proc_dointvec,
 496         },
 497         {
 498                 .procname       = "unaligned-trap",
 499                 .data           = &unaligned_enabled,
 500                 .maxlen         = sizeof (int),
 501                 .mode           = 0644,
 502                 .proc_handler   = proc_dointvec,
 503         },
 504 #endif
 505         {
 506                 .procname       = "ctrl-alt-del",
 507                 .data           = &C_A_D,
 508                 .maxlen         = sizeof(int),
 509                 .mode           = 0644,
 510                 .proc_handler   = proc_dointvec,
 511         },
 512 #ifdef CONFIG_FUNCTION_TRACER
 513         {
 514                 .procname       = "ftrace_enabled",
 515                 .data           = &ftrace_enabled,
 516                 .maxlen         = sizeof(int),
 517                 .mode           = 0644,
 518                 .proc_handler   = ftrace_enable_sysctl,
 519         },
 520 #endif
 521 #ifdef CONFIG_STACK_TRACER
 522         {
 523                 .procname       = "stack_tracer_enabled",
 524                 .data           = &stack_tracer_enabled,
 525                 .maxlen         = sizeof(int),
 526                 .mode           = 0644,
 527                 .proc_handler   = stack_trace_sysctl,
 528         },
 529 #endif
 530 #ifdef CONFIG_TRACING
 531         {
 532                 .procname       = "ftrace_dump_on_oops",
 533                 .data           = &ftrace_dump_on_oops,
 534                 .maxlen         = sizeof(int),
 535                 .mode           = 0644,
 536                 .proc_handler   = proc_dointvec,
 537         },
 538 #endif
 539 #ifdef CONFIG_MODULES
 540         {
 541                 .procname       = "modprobe",
 542                 .data           = &modprobe_path,
 543                 .maxlen         = KMOD_PATH_LEN,
 544                 .mode           = 0644,
 545                 .proc_handler   = proc_dostring,
 546         },
 547         {
 548                 .procname       = "modules_disabled",
 549                 .data           = &modules_disabled,
 550                 .maxlen         = sizeof(int),
 551                 .mode           = 0644,
 552                 /* only handle a transition from default "0" to "1" */
 553                 .proc_handler   = proc_dointvec_minmax,
 554                 .extra1         = &one,
 555                 .extra2         = &one,
 556         },
 557 #endif
 558 #if defined(CONFIG_HOTPLUG) && defined(CONFIG_NET)
 559         {
 560                 .procname       = "hotplug",
 561                 .data           = &uevent_helper,
 562                 .maxlen         = UEVENT_HELPER_PATH_LEN,
 563                 .mode           = 0644,
 564                 .proc_handler   = proc_dostring,
 565         },
 566 #endif
 567 #ifdef CONFIG_CHR_DEV_SG
 568         {
 569                 .procname       = "sg-big-buff",
 570                 .data           = &sg_big_buff,
 571                 .maxlen         = sizeof (int),
 572                 .mode           = 0444,
 573                 .proc_handler   = proc_dointvec,
 574         },
 575 #endif
 576 #ifdef CONFIG_BSD_PROCESS_ACCT
 577         {
 578                 .procname       = "acct",
 579                 .data           = &acct_parm,
 580                 .maxlen         = 3*sizeof(int),
 581                 .mode           = 0644,
 582                 .proc_handler   = proc_dointvec,
 583         },
 584 #endif
 585 #ifdef CONFIG_MAGIC_SYSRQ
 586         {
 587                 .procname       = "sysrq",
 588                 .data           = &__sysrq_enabled,
 589                 .maxlen         = sizeof (int),
 590                 .mode           = 0644,
 591                 .proc_handler   = sysrq_sysctl_handler,
 592         },
 593 #endif
 594 #ifdef CONFIG_PROC_SYSCTL
 595         {
 596                 .procname       = "cad_pid",
 597                 .data           = NULL,
 598                 .maxlen         = sizeof (int),
 599                 .mode           = 0600,
 600                 .proc_handler   = proc_do_cad_pid,
 601         },
 602 #endif
 603         {
 604                 .procname       = "threads-max",
 605                 .data           = &max_threads,
 606                 .maxlen         = sizeof(int),
 607                 .mode           = 0644,
 608                 .proc_handler   = proc_dointvec,
 609         },
 610         {
 611                 .procname       = "random",
 612                 .mode           = 0555,
 613                 .child          = random_table,
 614         },
 615         {
 616                 .procname       = "overflowuid",
 617                 .data           = &overflowuid,
 618                 .maxlen         = sizeof(int),
 619                 .mode           = 0644,
 620                 .proc_handler   = proc_dointvec_minmax,
 621                 .extra1         = &minolduid,
 622                 .extra2         = &maxolduid,
 623         },
 624         {
 625                 .procname       = "overflowgid",
 626                 .data           = &overflowgid,
 627                 .maxlen         = sizeof(int),
 628                 .mode           = 0644,
 629                 .proc_handler   = proc_dointvec_minmax,
 630                 .extra1         = &minolduid,
 631                 .extra2         = &maxolduid,
 632         },
 633 #ifdef CONFIG_S390
 634 #ifdef CONFIG_MATHEMU
 635         {
 636                 .procname       = "ieee_emulation_warnings",
 637                 .data           = &sysctl_ieee_emulation_warnings,
 638                 .maxlen         = sizeof(int),
 639                 .mode           = 0644,
 640                 .proc_handler   = proc_dointvec,
 641         },
 642 #endif
 643         {
 644                 .procname       = "userprocess_debug",
 645                 .data           = &show_unhandled_signals,
 646                 .maxlen         = sizeof(int),
 647                 .mode           = 0644,
 648                 .proc_handler   = proc_dointvec,
 649         },
 650 #endif
 651         {
 652                 .procname       = "pid_max",
 653                 .data           = &pid_max,
 654                 .maxlen         = sizeof (int),
 655                 .mode           = 0644,
 656                 .proc_handler   = proc_dointvec_minmax,
 657                 .extra1         = &pid_max_min,
 658                 .extra2         = &pid_max_max,
 659         },
 660         {
 661                 .procname       = "panic_on_oops",
 662                 .data           = &panic_on_oops,
 663                 .maxlen         = sizeof(int),
 664                 .mode           = 0644,
 665                 .proc_handler   = proc_dointvec,
 666         },
 667 #if defined CONFIG_PRINTK
 668         {
 669                 .procname       = "printk",
 670                 .data           = &console_loglevel,
 671                 .maxlen         = 4*sizeof(int),
 672                 .mode           = 0644,
 673                 .proc_handler   = proc_dointvec,
 674         },
 675         {
 676                 .procname       = "printk_ratelimit",
 677                 .data           = &printk_ratelimit_state.interval,
 678                 .maxlen         = sizeof(int),
 679                 .mode           = 0644,
 680                 .proc_handler   = proc_dointvec_jiffies,
 681         },
 682         {
 683                 .procname       = "printk_ratelimit_burst",
 684                 .data           = &printk_ratelimit_state.burst,
 685                 .maxlen         = sizeof(int),
 686                 .mode           = 0644,
 687                 .proc_handler   = proc_dointvec,
 688         },
 689         {
 690                 .procname       = "printk_delay",
 691                 .data           = &printk_delay_msec,
 692                 .maxlen         = sizeof(int),
 693                 .mode           = 0644,
 694                 .proc_handler   = proc_dointvec_minmax,
 695                 .extra1         = &zero,
 696                 .extra2         = &ten_thousand,
 697         },
 698 #endif
 699         {
 700                 .procname       = "ngroups_max",
 701                 .data           = &ngroups_max,
 702                 .maxlen         = sizeof (int),
 703                 .mode           = 0444,
 704                 .proc_handler   = proc_dointvec,
 705         },
 706 #if defined(CONFIG_X86_LOCAL_APIC) && defined(CONFIG_X86)
 707         {
 708                 .procname       = "unknown_nmi_panic",
 709                 .data           = &unknown_nmi_panic,
 710                 .maxlen         = sizeof (int),
 711                 .mode           = 0644,
 712                 .proc_handler   = proc_dointvec,
 713         },
 714         {
 715                 .procname       = "nmi_watchdog",
 716                 .data           = &nmi_watchdog_enabled,
 717                 .maxlen         = sizeof (int),
 718                 .mode           = 0644,
 719                 .proc_handler   = proc_nmi_enabled,
 720         },
 721 #endif
 722 #if defined(CONFIG_X86)
 723         {
 724                 .procname       = "panic_on_unrecovered_nmi",
 725                 .data           = &panic_on_unrecovered_nmi,
 726                 .maxlen         = sizeof(int),
 727                 .mode           = 0644,
 728                 .proc_handler   = proc_dointvec,
 729         },
 730         {
 731                 .procname       = "panic_on_io_nmi",
 732                 .data           = &panic_on_io_nmi,
 733                 .maxlen         = sizeof(int),
 734                 .mode           = 0644,
 735                 .proc_handler   = proc_dointvec,
 736         },
 737         {
 738                 .procname       = "bootloader_type",
 739                 .data           = &bootloader_type,
 740                 .maxlen         = sizeof (int),
 741                 .mode           = 0444,
 742                 .proc_handler   = proc_dointvec,
 743         },
 744         {
 745                 .procname       = "bootloader_version",
 746                 .data           = &bootloader_version,
 747                 .maxlen         = sizeof (int),
 748                 .mode           = 0444,
 749                 .proc_handler   = proc_dointvec,
 750         },
 751         {
 752                 .procname       = "kstack_depth_to_print",
 753                 .data           = &kstack_depth_to_print,
 754                 .maxlen         = sizeof(int),
 755                 .mode           = 0644,
 756                 .proc_handler   = proc_dointvec,
 757         },
 758         {
 759                 .procname       = "io_delay_type",
 760                 .data           = &io_delay_type,
 761                 .maxlen         = sizeof(int),
 762                 .mode           = 0644,
 763                 .proc_handler   = proc_dointvec,
 764         },
 765 #endif
 766 #if defined(CONFIG_MMU)
 767         {
 768                 .procname       = "randomize_va_space",
 769                 .data           = &randomize_va_space,
 770                 .maxlen         = sizeof(int),
 771                 .mode           = 0644,
 772                 .proc_handler   = proc_dointvec,
 773         },
 774 #endif
 775 #if defined(CONFIG_S390) && defined(CONFIG_SMP)
 776         {
 777                 .procname       = "spin_retry",
 778                 .data           = &spin_retry,
 779                 .maxlen         = sizeof (int),
 780                 .mode           = 0644,
 781                 .proc_handler   = proc_dointvec,
 782         },
 783 #endif
 784 #if     defined(CONFIG_ACPI_SLEEP) && defined(CONFIG_X86)
 785         {
 786                 .procname       = "acpi_video_flags",
 787                 .data           = &acpi_realmode_flags,
 788                 .maxlen         = sizeof (unsigned long),
 789                 .mode           = 0644,
 790                 .proc_handler   = proc_doulongvec_minmax,
 791         },
 792 #endif
 793 #ifdef CONFIG_IA64
 794         {
 795                 .procname       = "ignore-unaligned-usertrap",
 796                 .data           = &no_unaligned_warning,
 797                 .maxlen         = sizeof (int),
 798                 .mode           = 0644,
 799                 .proc_handler   = proc_dointvec,
 800         },
 801         {
 802                 .procname       = "unaligned-dump-stack",
 803                 .data           = &unaligned_dump_stack,
 804                 .maxlen         = sizeof (int),
 805                 .mode           = 0644,
 806                 .proc_handler   = proc_dointvec,
 807         },
 808 #endif
 809 #ifdef CONFIG_DETECT_SOFTLOCKUP
 810         {
 811                 .procname       = "softlockup_panic",
 812                 .data           = &softlockup_panic,
 813                 .maxlen         = sizeof(int),
 814                 .mode           = 0644,
 815                 .proc_handler   = proc_dointvec_minmax,
 816                 .extra1         = &zero,
 817                 .extra2         = &one,
 818         },
 819         {
 820                 .procname       = "softlockup_thresh",
 821                 .data           = &softlockup_thresh,
 822                 .maxlen         = sizeof(int),
 823                 .mode           = 0644,
 824                 .proc_handler   = proc_dosoftlockup_thresh,
 825                 .extra1         = &neg_one,
 826                 .extra2         = &sixty,
 827         },
 828 #endif
 829 #ifdef CONFIG_DETECT_HUNG_TASK
 830         {
 831                 .procname       = "hung_task_panic",
 832                 .data           = &sysctl_hung_task_panic,
 833                 .maxlen         = sizeof(int),
 834                 .mode           = 0644,
 835                 .proc_handler   = proc_dointvec_minmax,
 836                 .extra1         = &zero,
 837                 .extra2         = &one,
 838         },
 839         {
 840                 .procname       = "hung_task_check_count",
 841                 .data           = &sysctl_hung_task_check_count,
 842                 .maxlen         = sizeof(unsigned long),
 843                 .mode           = 0644,
 844                 .proc_handler   = proc_doulongvec_minmax,
 845         },
 846         {
 847                 .procname       = "hung_task_timeout_secs",
 848                 .data           = &sysctl_hung_task_timeout_secs,
 849                 .maxlen         = sizeof(unsigned long),
 850                 .mode           = 0644,
 851                 .proc_handler   = proc_dohung_task_timeout_secs,
 852         },
 853         {
 854                 .procname       = "hung_task_warnings",
 855                 .data           = &sysctl_hung_task_warnings,
 856                 .maxlen         = sizeof(unsigned long),
 857                 .mode           = 0644,
 858                 .proc_handler   = proc_doulongvec_minmax,
 859         },
 860 #endif
 861 #ifdef CONFIG_COMPAT
 862         {
 863                 .procname       = "compat-log",
 864                 .data           = &compat_log,
 865                 .maxlen         = sizeof (int),
 866                 .mode           = 0644,
 867                 .proc_handler   = proc_dointvec,
 868         },
 869 #endif
 870 #ifdef CONFIG_RT_MUTEXES
 871         {
 872                 .procname       = "max_lock_depth",
 873                 .data           = &max_lock_depth,
 874                 .maxlen         = sizeof(int),
 875                 .mode           = 0644,
 876                 .proc_handler   = proc_dointvec,
 877         },
 878 #endif
 879         {
 880                 .procname       = "poweroff_cmd",
 881                 .data           = &poweroff_cmd,
 882                 .maxlen         = POWEROFF_CMD_PATH_LEN,
 883                 .mode           = 0644,
 884                 .proc_handler   = proc_dostring,
 885         },
 886 #ifdef CONFIG_KEYS
 887         {
 888                 .procname       = "keys",
 889                 .mode           = 0555,
 890                 .child          = key_sysctls,
 891         },
 892 #endif
 893 #ifdef CONFIG_RCU_TORTURE_TEST
 894         {
 895                 .procname       = "rcutorture_runnable",
 896                 .data           = &rcutorture_runnable,
 897                 .maxlen         = sizeof(int),
 898                 .mode           = 0644,
 899                 .proc_handler   = proc_dointvec,
 900         },
 901 #endif
 902 #ifdef CONFIG_SLOW_WORK
 903         {
 904                 .procname       = "slow-work",
 905                 .mode           = 0555,
 906                 .child          = slow_work_sysctls,
 907         },
 908 #endif
 909 #ifdef CONFIG_PERF_EVENTS
 910         {
 911                 .procname       = "perf_event_paranoid",
 912                 .data           = &sysctl_perf_event_paranoid,
 913                 .maxlen         = sizeof(sysctl_perf_event_paranoid),
 914                 .mode           = 0644,
 915                 .proc_handler   = proc_dointvec,
 916         },
 917         {
 918                 .procname       = "perf_event_mlock_kb",
 919                 .data           = &sysctl_perf_event_mlock,
 920                 .maxlen         = sizeof(sysctl_perf_event_mlock),
 921                 .mode           = 0644,
 922                 .proc_handler   = proc_dointvec,
 923         },
 924         {
 925                 .procname       = "perf_event_max_sample_rate",
 926                 .data           = &sysctl_perf_event_sample_rate,
 927                 .maxlen         = sizeof(sysctl_perf_event_sample_rate),
 928                 .mode           = 0644,
 929                 .proc_handler   = proc_dointvec,
 930         },
 931 #endif
 932 #ifdef CONFIG_KMEMCHECK
 933         {
 934                 .procname       = "kmemcheck",
 935                 .data           = &kmemcheck_enabled,
 936                 .maxlen         = sizeof(int),
 937                 .mode           = 0644,
 938                 .proc_handler   = proc_dointvec,
 939         },
 940 #endif
 941 #ifdef CONFIG_BLOCK
 942         {
 943                 .procname       = "blk_iopoll",
 944                 .data           = &blk_iopoll_enabled,
 945                 .maxlen         = sizeof(int),
 946                 .mode           = 0644,
 947                 .proc_handler   = proc_dointvec,
 948         },
 949 #endif
 950 /*
 951  * NOTE: do not add new entries to this table unless you have read
 952  * Documentation/sysctl/ctl_unnumbered.txt
 953  */
 954         { }
 955 };
 956
 957 static struct ctl_table vm_table[] = {
 958         {
 959                 .procname       = "overcommit_memory",
 960                 .data           = &sysctl_overcommit_memory,
 961                 .maxlen         = sizeof(sysctl_overcommit_memory),
 962                 .mode           = 0644,
 963                 .proc_handler   = proc_dointvec,
 964         },
 965         {
 966                 .procname       = "panic_on_oom",
 967                 .data           = &sysctl_panic_on_oom,
 968                 .maxlen         = sizeof(sysctl_panic_on_oom),
 969                 .mode           = 0644,
 970                 .proc_handler   = proc_dointvec,
 971         },
 972         {
 973                 .procname       = "oom_kill_allocating_task",
 974                 .data           = &sysctl_oom_kill_allocating_task,
 975                 .maxlen         = sizeof(sysctl_oom_kill_allocating_task),
 976                 .mode           = 0644,
 977                 .proc_handler   = proc_dointvec,
 978         },
 979         {
 980                 .procname       = "oom_dump_tasks",
 981                 .data           = &sysctl_oom_dump_tasks,
 982                 .maxlen         = sizeof(sysctl_oom_dump_tasks),
 983                 .mode           = 0644,
 984                 .proc_handler   = proc_dointvec,
 985         },
 986         {
 987                 .procname       = "overcommit_ratio",
 988                 .data           = &sysctl_overcommit_ratio,
 989                 .maxlen         = sizeof(sysctl_overcommit_ratio),
 990                 .mode           = 0644,
 991                 .proc_handler   = proc_dointvec,
 992         },
 993         {
 994                 .procname       = "page-cluster",
 995                 .data           = &page_cluster,
 996                 .maxlen         = sizeof(int),
 997                 .mode           = 0644,
 998                 .proc_handler   = proc_dointvec,
 999         },
1000         {
1001                 .procname       = "dirty_background_ratio",
1002                 .data           = &dirty_background_ratio,
1003                 .maxlen         = sizeof(dirty_background_ratio),
1004                 .mode           = 0644,
1005                 .proc_handler   = dirty_background_ratio_handler,
1006                 .extra1         = &zero,
1007                 .extra2         = &one_hundred,
1008         },
1009         {
1010                 .procname       = "dirty_background_bytes",
1011                 .data           = &dirty_background_bytes,
1012                 .maxlen         = sizeof(dirty_background_bytes),
1013                 .mode           = 0644,
1014                 .proc_handler   = dirty_background_bytes_handler,
1015                 .extra1         = &one_ul,
1016         },
1017         {
1018                 .procname       = "dirty_ratio",
1019                 .data           = &vm_dirty_ratio,
1020                 .maxlen         = sizeof(vm_dirty_ratio),
1021                 .mode           = 0644,
1022                 .proc_handler   = dirty_ratio_handler,
1023                 .extra1         = &zero,
1024                 .extra2         = &one_hundred,
1025         },
1026         {
1027                 .procname       = "dirty_bytes",
1028                 .data           = &vm_dirty_bytes,
1029                 .maxlen         = sizeof(vm_dirty_bytes),
1030                 .mode           = 0644,
1031                 .proc_handler   = dirty_bytes_handler,
1032                 .extra1         = &dirty_bytes_min,
1033         },
1034         {
1035                 .procname       = "dirty_writeback_centisecs",
1036                 .data           = &dirty_writeback_interval,
1037                 .maxlen         = sizeof(dirty_writeback_interval),
1038                 .mode           = 0644,
1039                 .proc_handler   = dirty_writeback_centisecs_handler,
1040         },
1041         {
1042                 .procname       = "dirty_expire_centisecs",
1043                 .data           = &dirty_expire_interval,
1044                 .maxlen         = sizeof(dirty_expire_interval),
1045                 .mode           = 0644,
1046                 .proc_handler   = proc_dointvec,
1047         },
1048         {
1049                 .procname       = "nr_pdflush_threads",
1050                 .data           = &nr_pdflush_threads,
1051                 .maxlen         = sizeof nr_pdflush_threads,
1052                 .mode           = 0444 /* read-only*/,
1053                 .proc_handler   = proc_dointvec,
1054         },
1055         {
1056                 .procname       = "swappiness",
1057                 .data           = &vm_swappiness,
1058                 .maxlen         = sizeof(vm_swappiness),
1059                 .mode           = 0644,
1060                 .proc_handler   = proc_dointvec_minmax,
1061                 .extra1         = &zero,
1062                 .extra2         = &one_hundred,
1063         },
1064 #ifdef CONFIG_HUGETLB_PAGE
1065         {
1066                 .procname       = "nr_hugepages",
1067                 .data           = NULL,
1068                 .maxlen         = sizeof(unsigned long),
1069                 .mode           = 0644,
1070                 .proc_handler   = hugetlb_sysctl_handler,
1071                 .extra1         = (void *)&hugetlb_zero,
1072                 .extra2         = (void *)&hugetlb_infinity,
1073         },
1074 #ifdef CONFIG_NUMA
1075         {
1076                 .procname       = "nr_hugepages_mempolicy",
1077                 .data           = NULL,
1078                 .maxlen         = sizeof(unsigned long),
1079                 .mode           = 0644,
1080                 .proc_handler   = &hugetlb_mempolicy_sysctl_handler,
1081                 .extra1         = (void *)&hugetlb_zero,
1082                 .extra2         = (void *)&hugetlb_infinity,
1083         },
1084 #endif
1085          {
1086                 .procname       = "hugetlb_shm_group",
1087                 .data           = &sysctl_hugetlb_shm_group,
1088                 .maxlen         = sizeof(gid_t),
1089                 .mode           = 0644,
1090                 .proc_handler   = proc_dointvec,
1091          },
1092          {
1093                 .procname       = "hugepages_treat_as_movable",
1094                 .data           = &hugepages_treat_as_movable,
1095                 .maxlen         = sizeof(int),
1096                 .mode           = 0644,
1097                 .proc_handler   = hugetlb_treat_movable_handler,
1098         },
1099         {
1100                 .procname       = "nr_overcommit_hugepages",
1101                 .data           = NULL,
1102                 .maxlen         = sizeof(unsigned long),
1103                 .mode           = 0644,
1104                 .proc_handler   = hugetlb_overcommit_handler,
1105                 .extra1         = (void *)&hugetlb_zero,
1106                 .extra2         = (void *)&hugetlb_infinity,
1107         },
1108 #endif
1109         {
1110                 .procname       = "lowmem_reserve_ratio",
1111                 .data           = &sysctl_lowmem_reserve_ratio,
1112                 .maxlen         = sizeof(sysctl_lowmem_reserve_ratio),
1113                 .mode           = 0644,
1114                 .proc_handler   = lowmem_reserve_ratio_sysctl_handler,
1115         },
1116         {
1117                 .procname       = "drop_caches",
1118                 .data           = &sysctl_drop_caches,
1119                 .maxlen         = sizeof(int),
1120                 .mode           = 0644,
1121                 .proc_handler   = drop_caches_sysctl_handler,
1122         },
1123         {
1124                 .procname       = "min_free_kbytes",
1125                 .data           = &min_free_kbytes,
1126                 .maxlen         = sizeof(min_free_kbytes),
1127                 .mode           = 0644,
1128                 .proc_handler   = min_free_kbytes_sysctl_handler,
1129                 .extra1         = &zero,
1130         },
1131         {
1132                 .procname       = "percpu_pagelist_fraction",
1133                 .data           = &percpu_pagelist_fraction,
1134                 .maxlen         = sizeof(percpu_pagelist_fraction),
1135                 .mode           = 0644,
1136                 .proc_handler   = percpu_pagelist_fraction_sysctl_handler,
1137                 .extra1         = &min_percpu_pagelist_fract,
1138         },
1139 #ifdef CONFIG_MMU
1140         {
1141                 .procname       = "max_map_count",
1142                 .data           = &sysctl_max_map_count,
1143                 .maxlen         = sizeof(sysctl_max_map_count),
1144                 .mode           = 0644,
1145                 .proc_handler   = proc_dointvec_minmax,
1146                 .extra1         = &zero,
1147         },
1148 #else
1149         {
1150                 .procname       = "nr_trim_pages",
1151                 .data           = &sysctl_nr_trim_pages,
1152                 .maxlen         = sizeof(sysctl_nr_trim_pages),
1153                 .mode           = 0644,
1154                 .proc_handler   = proc_dointvec_minmax,
1155                 .extra1         = &zero,
1156         },
1157 #endif
1158         {
1159                 .procname       = "laptop_mode",
1160                 .data           = &laptop_mode,
1161                 .maxlen         = sizeof(laptop_mode),
1162                 .mode           = 0644,
1163                 .proc_handler   = proc_dointvec_jiffies,
1164         },
1165         {
1166                 .procname       = "block_dump",
1167                 .data           = &block_dump,
1168                 .maxlen         = sizeof(block_dump),
1169                 .mode           = 0644,
1170                 .proc_handler   = proc_dointvec,
1171                 .extra1         = &zero,
1172         },
1173         {
1174                 .procname       = "vfs_cache_pressure",
1175                 .data           = &sysctl_vfs_cache_pressure,
1176                 .maxlen         = sizeof(sysctl_vfs_cache_pressure),
1177                 .mode           = 0644,
1178                 .proc_handler   = proc_dointvec,
1179                 .extra1         = &zero,
1180         },
1181 #ifdef HAVE_ARCH_PICK_MMAP_LAYOUT
1182         {
1183                 .procname       = "legacy_va_layout",
1184                 .data           = &sysctl_legacy_va_layout,
1185                 .maxlen         = sizeof(sysctl_legacy_va_layout),
1186                 .mode           = 0644,
1187                 .proc_handler   = proc_dointvec,
1188                 .extra1         = &zero,
1189         },
1190 #endif
1191 #ifdef CONFIG_NUMA
1192         {
1193                 .procname       = "zone_reclaim_mode",
1194                 .data           = &zone_reclaim_mode,
1195                 .maxlen         = sizeof(zone_reclaim_mode),
1196                 .mode           = 0644,
1197                 .proc_handler   = proc_dointvec,
1198                 .extra1         = &zero,
1199         },
1200         {
1201                 .procname       = "min_unmapped_ratio",
1202                 .data           = &sysctl_min_unmapped_ratio,
1203                 .maxlen         = sizeof(sysctl_min_unmapped_ratio),
1204                 .mode           = 0644,
1205                 .proc_handler   = sysctl_min_unmapped_ratio_sysctl_handler,
1206                 .extra1         = &zero,
1207                 .extra2         = &one_hundred,
1208         },
1209         {
1210                 .procname       = "min_slab_ratio",
1211                 .data           = &sysctl_min_slab_ratio,
1212                 .maxlen         = sizeof(sysctl_min_slab_ratio),
1213                 .mode           = 0644,
1214                 .proc_handler   = sysctl_min_slab_ratio_sysctl_handler,
1215                 .extra1         = &zero,
1216                 .extra2         = &one_hundred,
1217         },
1218 #endif
1219 #ifdef CONFIG_SMP
1220         {
1221                 .procname       = "stat_interval",
1222                 .data           = &sysctl_stat_interval,
1223                 .maxlen         = sizeof(sysctl_stat_interval),
1224                 .mode           = 0644,
1225                 .proc_handler   = proc_dointvec_jiffies,
1226         },
1227 #endif
1228 #ifdef CONFIG_MMU
1229         {
1230                 .procname       = "mmap_min_addr",
1231                 .data           = &dac_mmap_min_addr,
1232                 .maxlen         = sizeof(unsigned long),
1233                 .mode           = 0644,
1234                 .proc_handler   = mmap_min_addr_handler,
1235         },
1236 #endif
1237 #ifdef CONFIG_NUMA
1238         {
1239                 .procname       = "numa_zonelist_order",
1240                 .data           = &numa_zonelist_order,
1241                 .maxlen         = NUMA_ZONELIST_ORDER_LEN,
1242                 .mode           = 0644,
1243                 .proc_handler   = numa_zonelist_order_handler,
1244         },
1245 #endif
1246 #if (defined(CONFIG_X86_32) && !defined(CONFIG_UML))|| \
1247    (defined(CONFIG_SUPERH) && defined(CONFIG_VSYSCALL))
1248         {
1249                 .procname       = "vdso_enabled",
1250                 .data           = &vdso_enabled,
1251                 .maxlen         = sizeof(vdso_enabled),
1252                 .mode           = 0644,
1253                 .proc_handler   = proc_dointvec,
1254                 .extra1         = &zero,
1255         },
1256 #endif
1257 #ifdef CONFIG_HIGHMEM
1258         {
1259                 .procname       = "highmem_is_dirtyable",
1260                 .data           = &vm_highmem_is_dirtyable,
1261                 .maxlen         = sizeof(vm_highmem_is_dirtyable),
1262                 .mode           = 0644,
1263                 .proc_handler   = proc_dointvec_minmax,
1264                 .extra1         = &zero,
1265                 .extra2         = &one,
1266         },
1267 #endif
1268         {
1269                 .procname       = "scan_unevictable_pages",
1270                 .data           = &scan_unevictable_pages,
1271                 .maxlen         = sizeof(scan_unevictable_pages),
1272                 .mode           = 0644,
1273                 .proc_handler   = scan_unevictable_handler,
1274         },
1275 #ifdef CONFIG_MEMORY_FAILURE
1276         {
1277                 .procname       = "memory_failure_early_kill",
1278                 .data           = &sysctl_memory_failure_early_kill,
1279                 .maxlen         = sizeof(sysctl_memory_failure_early_kill),
1280                 .mode           = 0644,
1281                 .proc_handler   = proc_dointvec_minmax,
1282                 .extra1         = &zero,
1283                 .extra2         = &one,
1284         },
1285         {
1286                 .procname       = "memory_failure_recovery",
1287                 .data           = &sysctl_memory_failure_recovery,
1288                 .maxlen         = sizeof(sysctl_memory_failure_recovery),
1289                 .mode           = 0644,
1290                 .proc_handler   = proc_dointvec_minmax,
1291                 .extra1         = &zero,
1292                 .extra2         = &one,
1293         },
1294 #endif
1295
1296 /*
1297  * NOTE: do not add new entries to this table unless you have read
1298  * Documentation/sysctl/ctl_unnumbered.txt
1299  */
1300         { }
1301 };
1302
1303 #if defined(CONFIG_BINFMT_MISC) || defined(CONFIG_BINFMT_MISC_MODULE)
1304 static struct ctl_table binfmt_misc_table[] = {
1305         { }
1306 };
1307 #endif
1308
1309 static struct ctl_table fs_table[] = {
1310         {
1311                 .procname       = "inode-nr",
1312                 .data           = &inodes_stat,
1313                 .maxlen         = 2*sizeof(int),
1314                 .mode           = 0444,
1315                 .proc_handler   = proc_dointvec,
1316         },
1317         {
1318                 .procname       = "inode-state",
1319                 .data           = &inodes_stat,
1320                 .maxlen         = 7*sizeof(int),
1321                 .mode           = 0444,
1322                 .proc_handler   = proc_dointvec,
1323         },
1324         {
1325                 .procname       = "file-nr",
1326                 .data           = &files_stat,
1327                 .maxlen         = 3*sizeof(int),
1328                 .mode           = 0444,
1329                 .proc_handler   = proc_nr_files,
1330         },
1331         {
1332                 .procname       = "file-max",
1333                 .data           = &files_stat.max_files,
1334                 .maxlen         = sizeof(int),
1335                 .mode           = 0644,
1336                 .proc_handler   = proc_dointvec,
1337         },
1338         {
1339                 .procname       = "nr_open",
1340                 .data           = &sysctl_nr_open,
1341                 .maxlen         = sizeof(int),
1342                 .mode           = 0644,
1343                 .proc_handler   = proc_dointvec_minmax,
1344                 .extra1         = &sysctl_nr_open_min,
1345                 .extra2         = &sysctl_nr_open_max,
1346         },
1347         {
1348                 .procname       = "dentry-state",
1349                 .data           = &dentry_stat,
1350                 .maxlen         = 6*sizeof(int),
1351                 .mode           = 0444,
1352                 .proc_handler   = proc_dointvec,
1353         },
1354         {
1355                 .procname       = "overflowuid",
1356                 .data           = &fs_overflowuid,
1357                 .maxlen         = sizeof(int),
1358                 .mode           = 0644,
1359                 .proc_handler   = proc_dointvec_minmax,
1360                 .extra1         = &minolduid,
1361                 .extra2         = &maxolduid,
1362         },
1363         {
1364                 .procname       = "overflowgid",
1365                 .data           = &fs_overflowgid,
1366                 .maxlen         = sizeof(int),
1367                 .mode           = 0644,
1368                 .proc_handler   = proc_dointvec_minmax,
1369                 .extra1         = &minolduid,
1370                 .extra2         = &maxolduid,
1371         },
1372 #ifdef CONFIG_FILE_LOCKING
1373         {
1374                 .procname       = "leases-enable",
1375                 .data           = &leases_enable,
1376                 .maxlen         = sizeof(int),
1377                 .mode           = 0644,
1378                 .proc_handler   = proc_dointvec,
1379         },
1380 #endif
1381 #ifdef CONFIG_DNOTIFY
1382         {
1383                 .procname       = "dir-notify-enable",
1384                 .data           = &dir_notify_enable,
1385                 .maxlen         = sizeof(int),
1386                 .mode           = 0644,
1387                 .proc_handler   = proc_dointvec,
1388         },
1389 #endif
1390 #ifdef CONFIG_MMU
1391 #ifdef CONFIG_FILE_LOCKING
1392         {
1393                 .procname       = "lease-break-time",
1394                 .data           = &lease_break_time,
1395                 .maxlen         = sizeof(int),
1396                 .mode           = 0644,
1397                 .proc_handler   = proc_dointvec,
1398         },
1399 #endif
1400 #ifdef CONFIG_AIO
1401         {
1402                 .procname       = "aio-nr",
1403                 .data           = &aio_nr,
1404                 .maxlen         = sizeof(aio_nr),
1405                 .mode           = 0444,
1406                 .proc_handler   = proc_doulongvec_minmax,
1407         },
1408         {
1409                 .procname       = "aio-max-nr",
1410                 .data           = &aio_max_nr,
1411                 .maxlen         = sizeof(aio_max_nr),
1412                 .mode           = 0644,
1413                 .proc_handler   = proc_doulongvec_minmax,
1414         },
1415 #endif /* CONFIG_AIO */
1416 #ifdef CONFIG_INOTIFY_USER
1417         {
1418                 .procname       = "inotify",
1419                 .mode           = 0555,
1420                 .child          = inotify_table,
1421         },
1422 #endif
1423 #ifdef CONFIG_EPOLL
1424         {
1425                 .procname       = "epoll",
1426                 .mode           = 0555,
1427                 .child          = epoll_table,
1428         },
1429 #endif
1430 #endif
1431         {
1432                 .procname       = "suid_dumpable",
1433                 .data           = &suid_dumpable,
1434                 .maxlen         = sizeof(int),
1435                 .mode           = 0644,
1436                 .proc_handler   = proc_dointvec_minmax,
1437                 .extra1         = &zero,
1438                 .extra2         = &two,
1439         },
1440 #if defined(CONFIG_BINFMT_MISC) || defined(CONFIG_BINFMT_MISC_MODULE)
1441         {
1442                 .procname       = "binfmt_misc",
1443                 .mode           = 0555,
1444                 .child          = binfmt_misc_table,
1445         },
1446 #endif
1447 /*
1448  * NOTE: do not add new entries to this table unless you have read
1449  * Documentation/sysctl/ctl_unnumbered.txt
1450  */
1451         { }
1452 };
1453
1454 static struct ctl_table debug_table[] = {
1455 #if defined(CONFIG_X86) || defined(CONFIG_PPC) || defined(CONFIG_SPARC) || \
1456     defined(CONFIG_S390)
1457         {
1458                 .procname       = "exception-trace",
1459                 .data           = &show_unhandled_signals,
1460                 .maxlen         = sizeof(int),
1461                 .mode           = 0644,
1462                 .proc_handler   = proc_dointvec
1463         },
1464 #endif
1465 #if defined(CONFIG_OPTPROBES)
1466         {
1467                 .procname       = "kprobes-optimization",
1468                 .data           = &sysctl_kprobes_optimization,
1469                 .maxlen         = sizeof(int),
1470                 .mode           = 0644,
1471                 .proc_handler   = proc_kprobes_optimization_handler,
1472                 .extra1         = &zero,
1473                 .extra2         = &one,
1474         },
1475 #endif
1476         { }
1477 };
1478
1479 static struct ctl_table dev_table[] = {
1480         { }
1481 };
1482
1483 static DEFINE_SPINLOCK(sysctl_lock);
1484
1485 /* called under sysctl_lock */
1486 static int use_table(struct ctl_table_header *p)
1487 {
1488         if (unlikely(p->unregistering))
1489                 return 0;
1490         p->used++;
1491         return 1;
1492 }
1493
1494 /* called under sysctl_lock */
1495 static void unuse_table(struct ctl_table_header *p)
1496 {
1497         if (!--p->used)
1498                 if (unlikely(p->unregistering))
1499                         complete(p->unregistering);
1500 }
1501
1502 /* called under sysctl_lock, will reacquire if has to wait */
1503 static void start_unregistering(struct ctl_table_header *p)
1504 {
1505         /*
1506          * if p->used is 0, nobody will ever touch that entry again;
1507          * we'll eliminate all paths to it before dropping sysctl_lock
1508          */
1509         if (unlikely(p->used)) {
1510                 struct completion wait;
1511                 init_completion(&wait);
1512                 p->unregistering = &wait;
1513                 spin_unlock(&sysctl_lock);
1514                 wait_for_completion(&wait);
1515                 spin_lock(&sysctl_lock);
1516         } else {
1517                 /* anything non-NULL; we'll never dereference it */
1518                 p->unregistering = ERR_PTR(-EINVAL);
1519         }
1520         /*
1521          * do not remove from the list until nobody holds it; walking the
1522          * list in do_sysctl() relies on that.
1523          */
1524         list_del_init(&p->ctl_entry);
1525 }
1526
1527 void sysctl_head_get(struct ctl_table_header *head)
1528 {
1529         spin_lock(&sysctl_lock);
1530         head->count++;
1531         spin_unlock(&sysctl_lock);
1532 }
1533
1534 void sysctl_head_put(struct ctl_table_header *head)
1535 {
1536         spin_lock(&sysctl_lock);
1537         if (!--head->count)
1538                 kfree(head);
1539         spin_unlock(&sysctl_lock);
1540 }
1541
1542 struct ctl_table_header *sysctl_head_grab(struct ctl_table_header *head)
1543 {
1544         if (!head)
1545                 BUG();
1546         spin_lock(&sysctl_lock);
1547         if (!use_table(head))
1548                 head = ERR_PTR(-ENOENT);
1549         spin_unlock(&sysctl_lock);
1550         return head;
1551 }
1552
1553 void sysctl_head_finish(struct ctl_table_header *head)
1554 {
1555         if (!head)
1556                 return;
1557         spin_lock(&sysctl_lock);
1558         unuse_table(head);
1559         spin_unlock(&sysctl_lock);
1560 }
1561
1562 static struct ctl_table_set *
1563 lookup_header_set(struct ctl_table_root *root, struct nsproxy *namespaces)
1564 {
1565         struct ctl_table_set *set = &root->default_set;
1566         if (root->lookup)
1567                 set = root->lookup(root, namespaces);
1568         return set;
1569 }
1570
1571 static struct list_head *
1572 lookup_header_list(struct ctl_table_root *root, struct nsproxy *namespaces)
1573 {
1574         struct ctl_table_set *set = lookup_header_set(root, namespaces);
1575         return &set->list;
1576 }
1577
1578 struct ctl_table_header *__sysctl_head_next(struct nsproxy *namespaces,
1579                                             struct ctl_table_header *prev)
1580 {
1581         struct ctl_table_root *root;
1582         struct list_head *header_list;
1583         struct ctl_table_header *head;
1584         struct list_head *tmp;
1585
1586         spin_lock(&sysctl_lock);
1587         if (prev) {
1588                 head = prev;
1589                 tmp = &prev->ctl_entry;
1590                 unuse_table(prev);
1591                 goto next;
1592         }
1593         tmp = &root_table_header.ctl_entry;
1594         for (;;) {
1595                 head = list_entry(tmp, struct ctl_table_header, ctl_entry);
1596
1597                 if (!use_table(head))
1598                         goto next;
1599                 spin_unlock(&sysctl_lock);
1600                 return head;
1601         next:
1602                 root = head->root;
1603                 tmp = tmp->next;
1604                 header_list = lookup_header_list(root, namespaces);
1605                 if (tmp != header_list)
1606                         continue;
1607
1608                 do {
1609                         root = list_entry(root->root_list.next,
1610                                         struct ctl_table_root, root_list);
1611                         if (root == &sysctl_table_root)
1612                                 goto out;
1613                         header_list = lookup_header_list(root, namespaces);
1614                 } while (list_empty(header_list));
1615                 tmp = header_list->next;
1616         }
1617 out:
1618         spin_unlock(&sysctl_lock);
1619         return NULL;
1620 }
1621
1622 struct ctl_table_header *sysctl_head_next(struct ctl_table_header *prev)
1623 {
1624         return __sysctl_head_next(current->nsproxy, prev);
1625 }
1626
1627 void register_sysctl_root(struct ctl_table_root *root)
1628 {
1629         spin_lock(&sysctl_lock);
1630         list_add_tail(&root->root_list, &sysctl_table_root.root_list);
1631         spin_unlock(&sysctl_lock);
1632 }
1633
1634 /*
1635  * sysctl_perm does NOT grant the superuser all rights automatically, because
1636  * some sysctl variables are readonly even to root.
1637  */
1638
1639 static int test_perm(int mode, int op)
1640 {
1641         if (!current_euid())
1642                 mode >>= 6;
1643         else if (in_egroup_p(0))
1644                 mode >>= 3;
1645         if ((op & ~mode & (MAY_READ|MAY_WRITE|MAY_EXEC)) == 0)
1646                 return 0;
1647         return -EACCES;
1648 }
1649
1650 int sysctl_perm(struct ctl_table_root *root, struct ctl_table *table, int op)
1651 {
1652         int error;
1653         int mode;
1654
1655         error = security_sysctl(table, op & (MAY_READ | MAY_WRITE | MAY_EXEC));
1656         if (error)
1657                 return error;
1658
1659         if (root->permissions)
1660                 mode = root->permissions(root, current->nsproxy, table);
1661         else
1662                 mode = table->mode;
1663
1664         return test_perm(mode, op);
1665 }
1666
1667 static void sysctl_set_parent(struct ctl_table *parent, struct ctl_table *table)
1668 {
1669         for (; table->procname; table++) {
1670                 table->parent = parent;
1671                 if (table->child)
1672                         sysctl_set_parent(table, table->child);
1673         }
1674 }
1675
1676 static __init int sysctl_init(void)
1677 {
1678         sysctl_set_parent(NULL, root_table);
1679 #ifdef CONFIG_SYSCTL_SYSCALL_CHECK
1680         {
1681                 int err;
1682                 err = sysctl_check_table(current->nsproxy, root_table);
1683         }
1684 #endif
1685         return 0;
1686 }
1687
1688 core_initcall(sysctl_init);
1689
1690 static struct ctl_table *is_branch_in(struct ctl_table *branch,
1691                                       struct ctl_table *table)
1692 {
1693         struct ctl_table *p;
1694         const char *s = branch->procname;
1695
1696         /* branch should have named subdirectory as its first element */
1697         if (!s || !branch->child)
1698                 return NULL;
1699
1700         /* ... and nothing else */
1701         if (branch[1].procname)
1702                 return NULL;
1703
1704         /* table should contain subdirectory with the same name */
1705         for (p = table; p->procname; p++) {
1706                 if (!p->child)
1707                         continue;
1708                 if (p->procname && strcmp(p->procname, s) == 0)
1709                         return p;
1710         }
1711         return NULL;
1712 }
1713
1714 /* see if attaching q to p would be an improvement */
1715 static void try_attach(struct ctl_table_header *p, struct ctl_table_header *q)
1716 {
1717         struct ctl_table *to = p->ctl_table, *by = q->ctl_table;
1718         struct ctl_table *next;
1719         int is_better = 0;
1720         int not_in_parent = !p->attached_by;
1721
1722         while ((next = is_branch_in(by, to)) != NULL) {
1723                 if (by == q->attached_by)
1724                         is_better = 1;
1725                 if (to == p->attached_by)
1726                         not_in_parent = 1;
1727                 by = by->child;
1728                 to = next->child;
1729         }
1730
1731         if (is_better && not_in_parent) {
1732                 q->attached_by = by;
1733                 q->attached_to = to;
1734                 q->parent = p;
1735         }
1736 }
1737
1738 /**
1739  * __register_sysctl_paths - register a sysctl hierarchy
1740  * @root: List of sysctl headers to register on
1741  * @namespaces: Data to compute which lists of sysctl entries are visible
1742  * @path: The path to the directory the sysctl table is in.
1743  * @table: the top-level table structure
1744  *
1745  * Register a sysctl table hierarchy. @table should be a filled in ctl_table
1746  * array. A completely 0 filled entry terminates the table.
1747  *
1748  * The members of the &struct ctl_table structure are used as follows:
1749  *
1750  * procname - the name of the sysctl file under /proc/sys. Set to %NULL to not
1751  *            enter a sysctl file
1752  *
1753  * data - a pointer to data for use by proc_handler
1754  *
1755  * maxlen - the maximum size in bytes of the data
1756  *
1757  * mode - the file permissions for the /proc/sys file, and for sysctl(2)
1758  *
1759  * child - a pointer to the child sysctl table if this entry is a directory, or
1760  *         %NULL.
1761  *
1762  * proc_handler - the text handler routine (described below)
1763  *
1764  * de - for internal use by the sysctl routines
1765  *
1766  * extra1, extra2 - extra pointers usable by the proc handler routines
1767  *
1768  * Leaf nodes in the sysctl tree will be represented by a single file
1769  * under /proc; non-leaf nodes will be represented by directories.
1770  *
1771  * sysctl(2) can automatically manage read and write requests through
1772  * the sysctl table.  The data and maxlen fields of the ctl_table
1773  * struct enable minimal validation of the values being written to be
1774  * performed, and the mode field allows minimal authentication.
1775  *
1776  * There must be a proc_handler routine for any terminal nodes
1777  * mirrored under /proc/sys (non-terminals are handled by a built-in
1778  * directory handler).  Several default handlers are available to
1779  * cover common cases -
1780  *
1781  * proc_dostring(), proc_dointvec(), proc_dointvec_jiffies(),
1782  * proc_dointvec_userhz_jiffies(), proc_dointvec_minmax(),
1783  * proc_doulongvec_ms_jiffies_minmax(), proc_doulongvec_minmax()
1784  *
1785  * It is the handler's job to read the input buffer from user memory
1786  * and process it. The handler should return 0 on success.
1787  *
1788  * This routine returns %NULL on a failure to register, and a pointer
1789  * to the table header on success.
1790  */
1791 struct ctl_table_header *__register_sysctl_paths(
1792         struct ctl_table_root *root,
1793         struct nsproxy *namespaces,
1794         const struct ctl_path *path, struct ctl_table *table)
1795 {
1796         struct ctl_table_header *header;
1797         struct ctl_table *new, **prevp;
1798         unsigned int n, npath;
1799         struct ctl_table_set *set;
1800
1801         /* Count the path components */
1802         for (npath = 0; path[npath].procname; ++npath)
1803                 ;
1804
1805         /*
1806          * For each path component, allocate a 2-element ctl_table array.
1807          * The first array element will be filled with the sysctl entry
1808          * for this, the second will be the sentinel (procname == 0).
1809          *
1810          * We allocate everything in one go so that we don't have to
1811          * worry about freeing additional memory in unregister_sysctl_table.
1812          */
1813         header = kzalloc(sizeof(struct ctl_table_header) +
1814                          (2 * npath * sizeof(struct ctl_table)), GFP_KERNEL);
1815         if (!header)
1816                 return NULL;
1817
1818         new = (struct ctl_table *) (header + 1);
1819
1820         /* Now connect the dots */
1821         prevp = &header->ctl_table;
1822         for (n = 0; n < npath; ++n, ++path) {
1823                 /* Copy the procname */
1824                 new->procname = path->procname;
1825                 new->mode     = 0555;
1826
1827                 *prevp = new;
1828                 prevp = &new->child;
1829
1830                 new += 2;
1831         }
1832         *prevp = table;
1833         header->ctl_table_arg = table;
1834
1835         INIT_LIST_HEAD(&header->ctl_entry);
1836         header->used = 0;
1837         header->unregistering = NULL;
1838         header->root = root;
1839         sysctl_set_parent(NULL, header->ctl_table);
1840         header->count = 1;
1841 #ifdef CONFIG_SYSCTL_SYSCALL_CHECK
1842         if (sysctl_check_table(namespaces, header->ctl_table)) {
1843                 kfree(header);
1844                 return NULL;
1845         }
1846 #endif
1847         spin_lock(&sysctl_lock);
1848         header->set = lookup_header_set(root, namespaces);
1849         header->attached_by = header->ctl_table;
1850         header->attached_to = root_table;
1851         header->parent = &root_table_header;
1852         for (set = header->set; set; set = set->parent) {
1853                 struct ctl_table_header *p;
1854                 list_for_each_entry(p, &set->list, ctl_entry) {
1855                         if (p->unregistering)
1856                                 continue;
1857                         try_attach(p, header);
1858                 }
1859         }
1860         header->parent->count++;
1861         list_add_tail(&header->ctl_entry, &header->set->list);
1862         spin_unlock(&sysctl_lock);
1863
1864         return header;
1865 }
1866
1867 /**
1868  * register_sysctl_table_path - register a sysctl table hierarchy
1869  * @path: The path to the directory the sysctl table is in.
1870  * @table: the top-level table structure
1871  *
1872  * Register a sysctl table hierarchy. @table should be a filled in ctl_table
1873  * array. A completely 0 filled entry terminates the table.
1874  *
1875  * See __register_sysctl_paths for more details.
1876  */
1877 struct ctl_table_header *register_sysctl_paths(const struct ctl_path *path,
1878                                                 struct ctl_table *table)
1879 {
1880         return __register_sysctl_paths(&sysctl_table_root, current->nsproxy,
1881                                         path, table);
1882 }
1883
1884 /**
1885  * register_sysctl_table - register a sysctl table hierarchy
1886  * @table: the top-level table structure
1887  *
1888  * Register a sysctl table hierarchy. @table should be a filled in ctl_table
1889  * array. A completely 0 filled entry terminates the table.
1890  *
1891  * See register_sysctl_paths for more details.
1892  */
1893 struct ctl_table_header *register_sysctl_table(struct ctl_table *table)
1894 {
1895         static const struct ctl_path null_path[] = { {} };
1896
1897         return register_sysctl_paths(null_path, table);
1898 }
1899
1900 /**
1901  * unregister_sysctl_table - unregister a sysctl table hierarchy
1902  * @header: the header returned from register_sysctl_table
1903  *
1904  * Unregisters the sysctl table and all children. proc entries may not
1905  * actually be removed until they are no longer used by anyone.
1906  */
1907 void unregister_sysctl_table(struct ctl_table_header * header)
1908 {
1909         might_sleep();
1910
1911         if (header == NULL)
1912                 return;
1913
1914         spin_lock(&sysctl_lock);
1915         start_unregistering(header);
1916         if (!--header->parent->count) {
1917                 WARN_ON(1);
1918                 kfree(header->parent);
1919         }
1920         if (!--header->count)
1921                 kfree(header);
1922         spin_unlock(&sysctl_lock);
1923 }
1924
1925 int sysctl_is_seen(struct ctl_table_header *p)
1926 {
1927         struct ctl_table_set *set = p->set;
1928         int res;
1929         spin_lock(&sysctl_lock);
1930         if (p->unregistering)
1931                 res = 0;
1932         else if (!set->is_seen)
1933                 res = 1;
1934         else
1935                 res = set->is_seen(set);
1936         spin_unlock(&sysctl_lock);
1937         return res;
1938 }
1939
1940 void setup_sysctl_set(struct ctl_table_set *p,
1941         struct ctl_table_set *parent,
1942         int (*is_seen)(struct ctl_table_set *))
1943 {
1944         INIT_LIST_HEAD(&p->list);
1945         p->parent = parent ? parent : &sysctl_table_root.default_set;
1946         p->is_seen = is_seen;
1947 }
1948
1949 #else /* !CONFIG_SYSCTL */
1950 struct ctl_table_header *register_sysctl_table(struct ctl_table * table)
1951 {
1952         return NULL;
1953 }
1954
1955 struct ctl_table_header *register_sysctl_paths(const struct ctl_path *path,
1956                                                     struct ctl_table *table)
1957 {
1958         return NULL;
1959 }
1960
1961 void unregister_sysctl_table(struct ctl_table_header * table)
1962 {
1963 }
1964
1965 void setup_sysctl_set(struct ctl_table_set *p,
1966         struct ctl_table_set *parent,
1967         int (*is_seen)(struct ctl_table_set *))
1968 {
1969 }
1970
1971 void sysctl_head_put(struct ctl_table_header *head)
1972 {
1973 }
1974
1975 #endif /* CONFIG_SYSCTL */
1976
1977 /*
1978  * /proc/sys support
1979  */
1980
1981 #ifdef CONFIG_PROC_SYSCTL
1982
1983 static int _proc_do_string(void* data, int maxlen, int write,
1984                            void __user *buffer,
1985                            size_t *lenp, loff_t *ppos)
1986 {
1987         size_t len;
1988         char __user *p;
1989         char c;
1990
1991         if (!data || !maxlen || !*lenp) {
1992                 *lenp = 0;
1993                 return 0;
1994         }
1995
1996         if (write) {
1997                 len = 0;
1998                 p = buffer;
1999                 while (len < *lenp) {
2000                         if (get_user(c, p++))
2001                                 return -EFAULT;
2002                         if (c == 0 || c == '\n')
2003                                 break;
2004                         len++;
2005                 }
2006                 if (len >= maxlen)
2007                         len = maxlen-1;
2008                 if(copy_from_user(data, buffer, len))
2009                         return -EFAULT;
2010                 ((char *) data)[len] = 0;
2011                 *ppos += *lenp;
2012         } else {
2013                 len = strlen(data);
2014                 if (len > maxlen)
2015                         len = maxlen;
2016
2017                 if (*ppos > len) {
2018                         *lenp = 0;
2019                         return 0;
2020                 }
2021
2022                 data += *ppos;
2023                 len  -= *ppos;
2024
2025                 if (len > *lenp)
2026                         len = *lenp;
2027                 if (len)
2028                         if(copy_to_user(buffer, data, len))
2029                                 return -EFAULT;
2030                 if (len < *lenp) {
2031                         if(put_user('\n', ((char __user *) buffer) + len))
2032                                 return -EFAULT;
2033                         len++;
2034                 }
2035                 *lenp = len;
2036                 *ppos += len;
2037         }
2038         return 0;
2039 }
2040
2041 /**
2042  * proc_dostring - read a string sysctl
2043  * @table: the sysctl table
2044  * @write: %TRUE if this is a write to the sysctl file
2045  * @buffer: the user buffer
2046  * @lenp: the size of the user buffer
2047  * @ppos: file position
2048  *
2049  * Reads/writes a string from/to the user buffer. If the kernel
2050  * buffer provided is not large enough to hold the string, the
2051  * string is truncated. The copied string is %NULL-terminated.
2052  * If the string is being read by the user process, it is copied
2053  * and a newline '\n' is added. It is truncated if the buffer is
2054  * not large enough.
2055  *
2056  * Returns 0 on success.
2057  */
2058 int proc_dostring(struct ctl_table *table, int write,
2059                   void __user *buffer, size_t *lenp, loff_t *ppos)
2060 {
2061         return _proc_do_string(table->data, table->maxlen, write,
2062                                buffer, lenp, ppos);
2063 }
2064
2065 static size_t proc_skip_spaces(char **buf)
2066 {
2067         size_t ret;
2068         char *tmp = skip_spaces(*buf);
2069         ret = tmp - *buf;
2070         *buf = tmp;
2071         return ret;
2072 }
2073
2074 static void proc_skip_char(char **buf, size_t *size, const char v)
2075 {
2076         while (*size) {
2077                 if (**buf != v)
2078                         break;
2079                 (*size)--;
2080                 (*buf)++;
2081         }
2082 }
2083
2084 #define TMPBUFLEN 22
2085 /**
2086  * proc_get_long - reads an ASCII formated integer from a user buffer
2087  *
2088  * @buf - a kernel buffer
2089  * @size - size of the kernel buffer
2090  * @val - this is where the number will be stored
2091  * @neg - set to %TRUE if number is negative
2092  * @perm_tr - a vector which contains the allowed trailers
2093  * @perm_tr_len - size of the perm_tr vector
2094  * @tr - pointer to store the trailer character
2095  *
2096  * In case of success 0 is returned and buf and size are updated with
2097  * the amount of bytes read. If tr is non NULL and a trailing
2098  * character exist (size is non zero after returning from this
2099  * function) tr is updated with the trailing character.
2100  */
2101 static int proc_get_long(char **buf, size_t *size,
2102                           unsigned long *val, bool *neg,
2103                           const char *perm_tr, unsigned perm_tr_len, char *tr)
2104 {
2105         int len;
2106         char *p, tmp[TMPBUFLEN];
2107
2108         if (!*size)
2109                 return -EINVAL;
2110
2111         len = *size;
2112         if (len > TMPBUFLEN - 1)
2113                 len = TMPBUFLEN - 1;
2114
2115         memcpy(tmp, *buf, len);
2116
2117         tmp[len] = 0;
2118         p = tmp;
2119         if (*p == '-' && *size > 1) {
2120                 *neg = true;
2121                 p++;
2122         } else
2123                 *neg = false;
2124         if (!isdigit(*p))
2125                 return -EINVAL;
2126
2127         *val = simple_strtoul(p, &p, 0);
2128
2129         len = p - tmp;
2130
2131         /* We don't know if the next char is whitespace thus we may accept
2132          * invalid integers (e.g. 1234...a) or two integers instead of one
2133          * (e.g. 123...1). So lets not allow such large numbers. */
2134         if (len == TMPBUFLEN - 1)
2135                 return -EINVAL;
2136
2137         if (len < *size && perm_tr_len && !memchr(perm_tr, *p, perm_tr_len))
2138                 return -EINVAL;
2139
2140         if (tr && (len < *size))
2141                 *tr = *p;
2142
2143         *buf += len;
2144         *size -= len;
2145
2146         return 0;
2147 }
2148
2149 /**
2150  * proc_put_long - coverts an integer to a decimal ASCII formated string
2151  *
2152  * @buf - the user buffer
2153  * @size - the size of the user buffer
2154  * @val - the integer to be converted
2155  * @neg - sign of the number, %TRUE for negative
2156  *
2157  * In case of success 0 is returned and buf and size are updated with
2158  * the amount of bytes read.
2159  */
2160 static int proc_put_long(void __user **buf, size_t *size, unsigned long val,
2161                           bool neg)
2162 {
2163         int len;
2164         char tmp[TMPBUFLEN], *p = tmp;
2165
2166         sprintf(p, "%s%lu", neg ? "-" : "", val);
2167         len = strlen(tmp);
2168         if (len > *size)
2169                 len = *size;
2170         if (copy_to_user(*buf, tmp, len))
2171                 return -EFAULT;
2172         *size -= len;
2173         *buf += len;
2174         return 0;
2175 }
2176 #undef TMPBUFLEN
2177
2178 static int proc_put_char(void __user **buf, size_t *size, char c)
2179 {
2180         if (*size) {
2181                 char __user **buffer = (char __user **)buf;
2182                 if (put_user(c, *buffer))
2183                         return -EFAULT;
2184                 (*size)--, (*buffer)++;
2185                 *buf = *buffer;
2186         }
2187         return 0;
2188 }
2189
2190 static int do_proc_dointvec_conv(bool *negp, unsigned long *lvalp,
2191                                  int *valp,
2192                                  int write, void *data)
2193 {
2194         if (write) {
2195                 *valp = *negp ? -*lvalp : *lvalp;
2196         } else {
2197                 int val = *valp;
2198                 if (val < 0) {
2199                         *negp = true;
2200                         *lvalp = (unsigned long)-val;
2201                 } else {
2202                         *negp = false;
2203                         *lvalp = (unsigned long)val;
2204                 }
2205         }
2206         return 0;
2207 }
2208
2209 static const char proc_wspace_sep[] = { ' ', '\t', '\n' };
2210
2211 static int __do_proc_dointvec(void *tbl_data, struct ctl_table *table,
2212                   int write, void __user *buffer,
2213                   size_t *lenp, loff_t *ppos,
2214                   int (*conv)(bool *negp, unsigned long *lvalp, int *valp,
2215                               int write, void *data),
2216                   void *data)
2217 {
2218         int *i, vleft, first = 1, err = 0;
2219         unsigned long page = 0;
2220         size_t left;
2221         char *kbuf;
2222
2223         if (!tbl_data || !table->maxlen || !*lenp || (*ppos && !write)) {
2224                 *lenp = 0;
2225                 return 0;
2226         }
2227
2228         i = (int *) tbl_data;
2229         vleft = table->maxlen / sizeof(*i);
2230         left = *lenp;
2231
2232         if (!conv)
2233                 conv = do_proc_dointvec_conv;
2234
2235         if (write) {
2236                 if (left > PAGE_SIZE - 1)
2237                         left = PAGE_SIZE - 1;
2238                 page = __get_free_page(GFP_TEMPORARY);
2239                 kbuf = (char *) page;
2240                 if (!kbuf)
2241                         return -ENOMEM;
2242                 if (copy_from_user(kbuf, buffer, left)) {
2243                         err = -EFAULT;
2244                         goto free;
2245                 }
2246                 kbuf[left] = 0;
2247         }
2248
2249         for (; left && vleft--; i++, first=0) {
2250                 unsigned long lval;
2251                 bool neg;
2252
2253                 if (write) {
2254                         left -= proc_skip_spaces(&kbuf);
2255
2256                         err = proc_get_long(&kbuf, &left, &lval, &neg,
2257                                              proc_wspace_sep,
2258                                              sizeof(proc_wspace_sep), NULL);
2259                         if (err)
2260                                 break;
2261                         if (conv(&neg, &lval, i, 1, data)) {
2262                                 err = -EINVAL;
2263                                 break;
2264                         }
2265                 } else {
2266                         if (conv(&neg, &lval, i, 0, data)) {
2267                                 err = -EINVAL;
2268                                 break;
2269                         }
2270                         if (!first)
2271                                 err = proc_put_char(&buffer, &left, '\t');
2272                         if (err)
2273                                 break;
2274                         err = proc_put_long(&buffer, &left, lval, neg);
2275                         if (err)
2276                                 break;
2277                 }
2278         }
2279
2280         if (!write && !first && left && !err)
2281                 err = proc_put_char(&buffer, &left, '\n');
2282         if (write && !err)
2283                 left -= proc_skip_spaces(&kbuf);
2284 free:
2285         if (write) {
2286                 free_page(page);
2287                 if (first)
2288                         return err ? : -EINVAL;
2289         }
2290         *lenp -= left;
2291         *ppos += *lenp;
2292         return err;
2293 }
2294
2295 static int do_proc_dointvec(struct ctl_table *table, int write,
2296                   void __user *buffer, size_t *lenp, loff_t *ppos,
2297                   int (*conv)(bool *negp, unsigned long *lvalp, int *valp,
2298                               int write, void *data),
2299                   void *data)
2300 {
2301         return __do_proc_dointvec(table->data, table, write,
2302                         buffer, lenp, ppos, conv, data);
2303 }
2304
2305 /**
2306  * proc_dointvec - read a vector of integers
2307  * @table: the sysctl table
2308  * @write: %TRUE if this is a write to the sysctl file
2309  * @buffer: the user buffer
2310  * @lenp: the size of the user buffer
2311  * @ppos: file position
2312  *
2313  * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2314  * values from/to the user buffer, treated as an ASCII string.
2315  *
2316  * Returns 0 on success.
2317  */
2318 int proc_dointvec(struct ctl_table *table, int write,
2319                      void __user *buffer, size_t *lenp, loff_t *ppos)
2320 {
2321     return do_proc_dointvec(table,write,buffer,lenp,ppos,
2322                             NULL,NULL);
2323 }
2324
2325 /*
2326  * Taint values can only be increased
2327  * This means we can safely use a temporary.
2328  */
2329 static int proc_taint(struct ctl_table *table, int write,
2330                                void __user *buffer, size_t *lenp, loff_t *ppos)
2331 {
2332         struct ctl_table t;
2333         unsigned long tmptaint = get_taint();
2334         int err;
2335
2336         if (write && !capable(CAP_SYS_ADMIN))
2337                 return -EPERM;
2338
2339         t = *table;
2340         t.data = &tmptaint;
2341         err = proc_doulongvec_minmax(&t, write, buffer, lenp, ppos);
2342         if (err < 0)
2343                 return err;
2344
2345         if (write) {
2346                 /*
2347                  * Poor man's atomic or. Not worth adding a primitive
2348                  * to everyone's atomic.h for this
2349                  */
2350                 int i;
2351                 for (i = 0; i < BITS_PER_LONG && tmptaint >> i; i++) {
2352                         if ((tmptaint >> i) & 1)
2353                                 add_taint(i);
2354                 }
2355         }
2356
2357         return err;
2358 }
2359
2360 struct do_proc_dointvec_minmax_conv_param {
2361         int *min;
2362         int *max;
2363 };
2364
2365 static int do_proc_dointvec_minmax_conv(bool *negp, unsigned long *lvalp,
2366                                         int *valp,
2367                                         int write, void *data)
2368 {
2369         struct do_proc_dointvec_minmax_conv_param *param = data;
2370         if (write) {
2371                 int val = *negp ? -*lvalp : *lvalp;
2372                 if ((param->min && *param->min > val) ||
2373                     (param->max && *param->max < val))
2374                         return -EINVAL;
2375                 *valp = val;
2376         } else {
2377                 int val = *valp;
2378                 if (val < 0) {
2379                         *negp = true;
2380                         *lvalp = (unsigned long)-val;
2381                 } else {
2382                         *negp = false;
2383                         *lvalp = (unsigned long)val;
2384                 }
2385         }
2386         return 0;
2387 }
2388
2389 /**
2390  * proc_dointvec_minmax - read a vector of integers with min/max values
2391  * @table: the sysctl table
2392  * @write: %TRUE if this is a write to the sysctl file
2393  * @buffer: the user buffer
2394  * @lenp: the size of the user buffer
2395  * @ppos: file position
2396  *
2397  * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2398  * values from/to the user buffer, treated as an ASCII string.
2399  *
2400  * This routine will ensure the values are within the range specified by
2401  * table->extra1 (min) and table->extra2 (max).
2402  *
2403  * Returns 0 on success.
2404  */
2405 int proc_dointvec_minmax(struct ctl_table *table, int write,
2406                   void __user *buffer, size_t *lenp, loff_t *ppos)
2407 {
2408         struct do_proc_dointvec_minmax_conv_param param = {
2409                 .min = (int *) table->extra1,
2410                 .max = (int *) table->extra2,
2411         };
2412         return do_proc_dointvec(table, write, buffer, lenp, ppos,
2413                                 do_proc_dointvec_minmax_conv, &param);
2414 }
2415
2416 static int __do_proc_doulongvec_minmax(void *data, struct ctl_table *table, int write,
2417                                      void __user *buffer,
2418                                      size_t *lenp, loff_t *ppos,
2419                                      unsigned long convmul,
2420                                      unsigned long convdiv)
2421 {
2422         unsigned long *i, *min, *max;
2423         int vleft, first = 1, err = 0;
2424         unsigned long page = 0;
2425         size_t left;
2426         char *kbuf;
2427
2428         if (!data || !table->maxlen || !*lenp || (*ppos && !write)) {
2429                 *lenp = 0;
2430                 return 0;
2431         }
2432
2433         i = (unsigned long *) data;
2434         min = (unsigned long *) table->extra1;
2435         max = (unsigned long *) table->extra2;
2436         vleft = table->maxlen / sizeof(unsigned long);
2437         left = *lenp;
2438
2439         if (write) {
2440                 if (left > PAGE_SIZE - 1)
2441                         left = PAGE_SIZE - 1;
2442                 page = __get_free_page(GFP_TEMPORARY);
2443                 kbuf = (char *) page;
2444                 if (!kbuf)
2445                         return -ENOMEM;
2446                 if (copy_from_user(kbuf, buffer, left)) {
2447                         err = -EFAULT;
2448                         goto free;
2449                 }
2450                 kbuf[left] = 0;
2451         }
2452
2453         for (; left && vleft--; i++, min++, max++, first=0) {
2454                 unsigned long val;
2455
2456                 if (write) {
2457                         bool neg;
2458
2459                         left -= proc_skip_spaces(&kbuf);
2460
2461                         err = proc_get_long(&kbuf, &left, &val, &neg,
2462                                              proc_wspace_sep,
2463                                              sizeof(proc_wspace_sep), NULL);
2464                         if (err)
2465                                 break;
2466                         if (neg)
2467                                 continue;
2468                         if ((min && val < *min) || (max && val > *max))
2469                                 continue;
2470                         *i = val;
2471                 } else {
2472                         val = convdiv * (*i) / convmul;
2473                         if (!first)
2474                                 err = proc_put_char(&buffer, &left, '\t');
2475                         err = proc_put_long(&buffer, &left, val, false);
2476                         if (err)
2477                                 break;
2478                 }
2479         }
2480
2481         if (!write && !first && left && !err)
2482                 err = proc_put_char(&buffer, &left, '\n');
2483         if (write && !err)
2484                 left -= proc_skip_spaces(&kbuf);
2485 free:
2486         if (write) {
2487                 free_page(page);
2488                 if (first)
2489                         return err ? : -EINVAL;
2490         }
2491         *lenp -= left;
2492         *ppos += *lenp;
2493         return err;
2494 }
2495
2496 static int do_proc_doulongvec_minmax(struct ctl_table *table, int write,
2497                                      void __user *buffer,
2498                                      size_t *lenp, loff_t *ppos,
2499                                      unsigned long convmul,
2500                                      unsigned long convdiv)
2501 {
2502         return __do_proc_doulongvec_minmax(table->data, table, write,
2503                         buffer, lenp, ppos, convmul, convdiv);
2504 }
2505
2506 /**
2507  * proc_doulongvec_minmax - read a vector of long integers with min/max values
2508  * @table: the sysctl table
2509  * @write: %TRUE if this is a write to the sysctl file
2510  * @buffer: the user buffer
2511  * @lenp: the size of the user buffer
2512  * @ppos: file position
2513  *
2514  * Reads/writes up to table->maxlen/sizeof(unsigned long) unsigned long
2515  * values from/to the user buffer, treated as an ASCII string.
2516  *
2517  * This routine will ensure the values are within the range specified by
2518  * table->extra1 (min) and table->extra2 (max).
2519  *
2520  * Returns 0 on success.
2521  */
2522 int proc_doulongvec_minmax(struct ctl_table *table, int write,
2523                            void __user *buffer, size_t *lenp, loff_t *ppos)
2524 {
2525     return do_proc_doulongvec_minmax(table, write, buffer, lenp, ppos, 1l, 1l);
2526 }
2527
2528 /**
2529  * proc_doulongvec_ms_jiffies_minmax - read a vector of millisecond values with min/max values
2530  * @table: the sysctl table
2531  * @write: %TRUE if this is a write to the sysctl file
2532  * @buffer: the user buffer
2533  * @lenp: the size of the user buffer
2534  * @ppos: file position
2535  *
2536  * Reads/writes up to table->maxlen/sizeof(unsigned long) unsigned long
2537  * values from/to the user buffer, treated as an ASCII string. The values
2538  * are treated as milliseconds, and converted to jiffies when they are stored.
2539  *
2540  * This routine will ensure the values are within the range specified by
2541  * table->extra1 (min) and table->extra2 (max).
2542  *
2543  * Returns 0 on success.
2544  */
2545 int proc_doulongvec_ms_jiffies_minmax(struct ctl_table *table, int write,
2546                                       void __user *buffer,
2547                                       size_t *lenp, loff_t *ppos)
2548 {
2549     return do_proc_doulongvec_minmax(table, write, buffer,
2550                                      lenp, ppos, HZ, 1000l);
2551 }
2552
2553
2554 static int do_proc_dointvec_jiffies_conv(bool *negp, unsigned long *lvalp,
2555                                          int *valp,
2556                                          int write, void *data)
2557 {
2558         if (write) {
2559                 if (*lvalp > LONG_MAX / HZ)
2560                         return 1;
2561                 *valp = *negp ? -(*lvalp*HZ) : (*lvalp*HZ);
2562         } else {
2563                 int val = *valp;
2564                 unsigned long lval;
2565                 if (val < 0) {
2566                         *negp = true;
2567                         lval = (unsigned long)-val;
2568                 } else {
2569                         *negp = false;
2570                         lval = (unsigned long)val;
2571                 }
2572                 *lvalp = lval / HZ;
2573         }
2574         return 0;
2575 }
2576
2577 static int do_proc_dointvec_userhz_jiffies_conv(bool *negp, unsigned long *lvalp,
2578                                                 int *valp,
2579                                                 int write, void *data)
2580 {
2581         if (write) {
2582                 if (USER_HZ < HZ && *lvalp > (LONG_MAX / HZ) * USER_HZ)
2583                         return 1;
2584                 *valp = clock_t_to_jiffies(*negp ? -*lvalp : *lvalp);
2585         } else {
2586                 int val = *valp;
2587                 unsigned long lval;
2588                 if (val < 0) {
2589                         *negp = true;
2590                         lval = (unsigned long)-val;
2591                 } else {
2592                         *negp = false;
2593                         lval = (unsigned long)val;
2594                 }
2595                 *lvalp = jiffies_to_clock_t(lval);
2596         }
2597         return 0;
2598 }
2599
2600 static int do_proc_dointvec_ms_jiffies_conv(bool *negp, unsigned long *lvalp,
2601                                             int *valp,
2602                                             int write, void *data)
2603 {
2604         if (write) {
2605                 *valp = msecs_to_jiffies(*negp ? -*lvalp : *lvalp);
2606         } else {
2607                 int val = *valp;
2608                 unsigned long lval;
2609                 if (val < 0) {
2610                         *negp = true;
2611                         lval = (unsigned long)-val;
2612                 } else {
2613                         *negp = false;
2614                         lval = (unsigned long)val;
2615                 }
2616                 *lvalp = jiffies_to_msecs(lval);
2617         }
2618         return 0;
2619 }
2620
2621 /**
2622  * proc_dointvec_jiffies - read a vector of integers as seconds
2623  * @table: the sysctl table
2624  * @write: %TRUE if this is a write to the sysctl file
2625  * @buffer: the user buffer
2626  * @lenp: the size of the user buffer
2627  * @ppos: file position
2628  *
2629  * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2630  * values from/to the user buffer, treated as an ASCII string.
2631  * The values read are assumed to be in seconds, and are converted into
2632  * jiffies.
2633  *
2634  * Returns 0 on success.
2635  */
2636 int proc_dointvec_jiffies(struct ctl_table *table, int write,
2637                           void __user *buffer, size_t *lenp, loff_t *ppos)
2638 {
2639     return do_proc_dointvec(table,write,buffer,lenp,ppos,
2640                             do_proc_dointvec_jiffies_conv,NULL);
2641 }
2642
2643 /**
2644  * proc_dointvec_userhz_jiffies - read a vector of integers as 1/USER_HZ seconds
2645  * @table: the sysctl table
2646  * @write: %TRUE if this is a write to the sysctl file
2647  * @buffer: the user buffer
2648  * @lenp: the size of the user buffer
2649  * @ppos: pointer to the file position
2650  *
2651  * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2652  * values from/to the user buffer, treated as an ASCII string.
2653  * The values read are assumed to be in 1/USER_HZ seconds, and
2654  * are converted into jiffies.
2655  *
2656  * Returns 0 on success.
2657  */
2658 int proc_dointvec_userhz_jiffies(struct ctl_table *table, int write,
2659                                  void __user *buffer, size_t *lenp, loff_t *ppos)
2660 {
2661     return do_proc_dointvec(table,write,buffer,lenp,ppos,
2662                             do_proc_dointvec_userhz_jiffies_conv,NULL);
2663 }
2664
2665 /**
2666  * proc_dointvec_ms_jiffies - read a vector of integers as 1 milliseconds
2667  * @table: the sysctl table
2668  * @write: %TRUE if this is a write to the sysctl file
2669  * @buffer: the user buffer
2670  * @lenp: the size of the user buffer
2671  * @ppos: file position
2672  * @ppos: the current position in the file
2673  *
2674  * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2675  * values from/to the user buffer, treated as an ASCII string.
2676  * The values read are assumed to be in 1/1000 seconds, and
2677  * are converted into jiffies.
2678  *
2679  * Returns 0 on success.
2680  */
2681 int proc_dointvec_ms_jiffies(struct ctl_table *table, int write,
2682                              void __user *buffer, size_t *lenp, loff_t *ppos)
2683 {
2684         return do_proc_dointvec(table, write, buffer, lenp, ppos,
2685                                 do_proc_dointvec_ms_jiffies_conv, NULL);
2686 }
2687
2688 static int proc_do_cad_pid(struct ctl_table *table, int write,
2689                            void __user *buffer, size_t *lenp, loff_t *ppos)
2690 {
2691         struct pid *new_pid;
2692         pid_t tmp;
2693         int r;
2694
2695         tmp = pid_vnr(cad_pid);
2696
2697         r = __do_proc_dointvec(&tmp, table, write, buffer,
2698                                lenp, ppos, NULL, NULL);
2699         if (r || !write)
2700                 return r;
2701
2702         new_pid = find_get_pid(tmp);
2703         if (!new_pid)
2704                 return -ESRCH;
2705
2706         put_pid(xchg(&cad_pid, new_pid));
2707         return 0;
2708 }
2709
2710 /**
2711  * proc_do_large_bitmap - read/write from/to a large bitmap
2712  * @table: the sysctl table
2713  * @write: %TRUE if this is a write to the sysctl file
2714  * @buffer: the user buffer
2715  * @lenp: the size of the user buffer
2716  * @ppos: file position
2717  *
2718  * The bitmap is stored at table->data and the bitmap length (in bits)
2719  * in table->maxlen.
2720  *
2721  * We use a range comma separated format (e.g. 1,3-4,10-10) so that
2722  * large bitmaps may be represented in a compact manner. Writing into
2723  * the file will clear the bitmap then update it with the given input.
2724  *
2725  * Returns 0 on success.
2726  */
2727 int proc_do_large_bitmap(struct ctl_table *table, int write,
2728                          void __user *buffer, size_t *lenp, loff_t *ppos)
2729 {
2730         int err = 0;
2731         bool first = 1;
2732         size_t left = *lenp;
2733         unsigned long bitmap_len = table->maxlen;
2734         unsigned long *bitmap = (unsigned long *) table->data;
2735         unsigned long *tmp_bitmap = NULL;
2736         char tr_a[] = { '-', ',', '\n' }, tr_b[] = { ',', '\n', 0 }, c;
2737
2738         if (!bitmap_len || !left || (*ppos && !write)) {
2739                 *lenp = 0;
2740                 return 0;
2741         }
2742
2743         if (write) {
2744                 unsigned long page = 0;
2745                 char *kbuf;
2746
2747                 if (left > PAGE_SIZE - 1)
2748                         left = PAGE_SIZE - 1;
2749
2750                 page = __get_free_page(GFP_TEMPORARY);
2751                 kbuf = (char *) page;
2752                 if (!kbuf)
2753                         return -ENOMEM;
2754                 if (copy_from_user(kbuf, buffer, left)) {
2755                         free_page(page);
2756                         return -EFAULT;
2757                 }
2758                 kbuf[left] = 0;
2759
2760                 tmp_bitmap = kzalloc(BITS_TO_LONGS(bitmap_len) * sizeof(unsigned long),
2761                                      GFP_KERNEL);
2762                 if (!tmp_bitmap) {
2763                         free_page(page);
2764                         return -ENOMEM;
2765                 }
2766                 proc_skip_char(&kbuf, &left, '\n');
2767                 while (!err && left) {
2768                         unsigned long val_a, val_b;
2769                         bool neg;
2770
2771                         err = proc_get_long(&kbuf, &left, &val_a, &neg, tr_a,
2772                                              sizeof(tr_a), &c);
2773                         if (err)
2774                                 break;
2775                         if (val_a >= bitmap_len || neg) {
2776                                 err = -EINVAL;
2777                                 break;
2778                         }
2779
2780                         val_b = val_a;
2781                         if (left) {
2782                                 kbuf++;
2783                                 left--;
2784                         }
2785
2786                         if (c == '-') {
2787                                 err = proc_get_long(&kbuf, &left, &val_b,
2788                                                      &neg, tr_b, sizeof(tr_b),
2789                                                      &c);
2790                                 if (err)
2791                                         break;
2792                                 if (val_b >= bitmap_len || neg ||
2793                                     val_a > val_b) {
2794                                         err = -EINVAL;
2795                                         break;
2796                                 }
2797                                 if (left) {
2798                                         kbuf++;
2799                                         left--;
2800                                 }
2801                         }
2802
2803                         while (val_a <= val_b)
2804                                 set_bit(val_a++, tmp_bitmap);
2805
2806                         first = 0;
2807                         proc_skip_char(&kbuf, &left, '\n');
2808                 }
2809                 free_page(page);
2810         } else {
2811                 unsigned long bit_a, bit_b = 0;
2812
2813                 while (left) {
2814                         bit_a = find_next_bit(bitmap, bitmap_len, bit_b);
2815                         if (bit_a >= bitmap_len)
2816                                 break;
2817                         bit_b = find_next_zero_bit(bitmap, bitmap_len,
2818                                                    bit_a + 1) - 1;
2819
2820                         if (!first) {
2821                                 err = proc_put_char(&buffer, &left, ',');
2822                                 if (err)
2823                                         break;
2824                         }
2825                         err = proc_put_long(&buffer, &left, bit_a, false);
2826                         if (err)
2827                                 break;
2828                         if (bit_a != bit_b) {
2829                                 err = proc_put_char(&buffer, &left, '-');
2830                                 if (err)
2831                                         break;
2832                                 err = proc_put_long(&buffer, &left, bit_b, false);
2833                                 if (err)
2834                                         break;
2835                         }
2836
2837                         first = 0; bit_b++;
2838                 }
2839                 if (!err)
2840                         err = proc_put_char(&buffer, &left, '\n');
2841         }
2842
2843         if (!err) {
2844                 if (write) {
2845                         if (*ppos)
2846                                 bitmap_or(bitmap, bitmap, tmp_bitmap, bitmap_len);
2847                         else
2848                                 memcpy(bitmap, tmp_bitmap,
2849                                         BITS_TO_LONGS(bitmap_len) * sizeof(unsigned long));
2850                 }
2851                 kfree(tmp_bitmap);
2852                 *lenp -= left;
2853                 *ppos += *lenp;
2854                 return 0;
2855         } else {
2856                 kfree(tmp_bitmap);
2857                 return err;
2858         }
2859 }
2860
2861 #else /* CONFIG_PROC_FS */
2862
2863 int proc_dostring(struct ctl_table *table, int write,
2864                   void __user *buffer, size_t *lenp, loff_t *ppos)
2865 {
2866         return -ENOSYS;
2867 }
2868
2869 int proc_dointvec(struct ctl_table *table, int write,
2870                   void __user *buffer, size_t *lenp, loff_t *ppos)
2871 {
2872         return -ENOSYS;
2873 }
2874
2875 int proc_dointvec_minmax(struct ctl_table *table, int write,
2876                     void __user *buffer, size_t *lenp, loff_t *ppos)
2877 {
2878         return -ENOSYS;
2879 }
2880
2881 int proc_dointvec_jiffies(struct ctl_table *table, int write,
2882                     void __user *buffer, size_t *lenp, loff_t *ppos)
2883 {
2884         return -ENOSYS;
2885 }
2886
2887 int proc_dointvec_userhz_jiffies(struct ctl_table *table, int write,
2888                     void __user *buffer, size_t *lenp, loff_t *ppos)
2889 {
2890         return -ENOSYS;
2891 }
2892
2893 int proc_dointvec_ms_jiffies(struct ctl_table *table, int write,
2894                              void __user *buffer, size_t *lenp, loff_t *ppos)
2895 {
2896         return -ENOSYS;
2897 }
2898
2899 int proc_doulongvec_minmax(struct ctl_table *table, int write,
2900                     void __user *buffer, size_t *lenp, loff_t *ppos)
2901 {
2902         return -ENOSYS;
2903 }
2904
2905 int proc_doulongvec_ms_jiffies_minmax(struct ctl_table *table, int write,
2906                                       void __user *buffer,
2907                                       size_t *lenp, loff_t *ppos)
2908 {
2909     return -ENOSYS;
2910 }
2911
2912
2913 #endif /* CONFIG_PROC_FS */
2914
2915 /*
2916  * No sense putting this after each symbol definition, twice,
2917  * exception granted :-)
2918  */
2919 EXPORT_SYMBOL(proc_dointvec);
2920 EXPORT_SYMBOL(proc_dointvec_jiffies);
2921 EXPORT_SYMBOL(proc_dointvec_minmax);
2922 EXPORT_SYMBOL(proc_dointvec_userhz_jiffies);
2923 EXPORT_SYMBOL(proc_dointvec_ms_jiffies);
2924 EXPORT_SYMBOL(proc_dostring);
2925 EXPORT_SYMBOL(proc_doulongvec_minmax);
2926 EXPORT_SYMBOL(proc_doulongvec_ms_jiffies_minmax);
2927 EXPORT_SYMBOL(register_sysctl_table);
2928 EXPORT_SYMBOL(register_sysctl_paths);
2929 EXPORT_SYMBOL(unregister_sysctl_table);