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/compaction.h>
  41 #include <linux/hugetlb.h>
  42 #include <linux/initrd.h>
  43 #include <linux/key.h>
  44 #include <linux/times.h>
  45 #include <linux/limits.h>
  46 #include <linux/dcache.h>
  47 #include <linux/dnotify.h>
  48 #include <linux/syscalls.h>
  49 #include <linux/vmstat.h>
  50 #include <linux/nfs_fs.h>
  51 #include <linux/acpi.h>
  52 #include <linux/reboot.h>
  53 #include <linux/ftrace.h>
  54 #include <linux/perf_event.h>
  55 #include <linux/kprobes.h>
  56 #include <linux/pipe_fs_i.h>
  57 #include <linux/oom.h>
  58
  59 #include <asm/uaccess.h>
  60 #include <asm/processor.h>
  61
  62 #ifdef CONFIG_X86
  63 #include <asm/nmi.h>
  64 #include <asm/stacktrace.h>
  65 #include <asm/io.h>
  66 #endif
  67 #ifdef CONFIG_BSD_PROCESS_ACCT
  68 #include <linux/acct.h>
  69 #endif
  70 #ifdef CONFIG_RT_MUTEXES
  71 #include <linux/rtmutex.h>
  72 #endif
  73 #if defined(CONFIG_PROVE_LOCKING) || defined(CONFIG_LOCK_STAT)
  74 #include <linux/lockdep.h>
  75 #endif
  76 #ifdef CONFIG_CHR_DEV_SG
  77 #include <scsi/sg.h>
  78 #endif
  79
  80 #ifdef CONFIG_LOCKUP_DETECTOR
  81 #include <linux/nmi.h>
  82 #endif
  83
  84
  85 #if defined(CONFIG_SYSCTL)
  86
  87 /* External variables not in a header file. */
  88 extern int sysctl_overcommit_memory;
  89 extern int sysctl_overcommit_ratio;
  90 extern int max_threads;
  91 extern int core_uses_pid;
  92 extern int suid_dumpable;
  93 extern char core_pattern[];
  94 extern unsigned int core_pipe_limit;
  95 extern int pid_max;
  96 extern int min_free_kbytes;
  97 extern int pid_max_min, pid_max_max;
  98 extern int sysctl_drop_caches;
  99 extern int percpu_pagelist_fraction;
 100 extern int compat_log;
 101 extern int latencytop_enabled;
 102 extern int sysctl_nr_open_min, sysctl_nr_open_max;
 103 #ifndef CONFIG_MMU
 104 extern int sysctl_nr_trim_pages;
 105 #endif
 106 #ifdef CONFIG_BLOCK
 107 extern int blk_iopoll_enabled;
 108 #endif
 109
 110 /* Constants used for minimum and  maximum */
 111 #ifdef CONFIG_LOCKUP_DETECTOR
 112 static int sixty = 60;
 113 static int neg_one = -1;
 114 #endif
 115
 116 static int zero;
 117 static int __maybe_unused one = 1;
 118 static int __maybe_unused two = 2;
 119 static unsigned long one_ul = 1;
 120 static int one_hundred = 100;
 121 #ifdef CONFIG_PRINTK
 122 static int ten_thousand = 10000;
 123 #endif
 124
 125 /* this is needed for the proc_doulongvec_minmax of vm_dirty_bytes */
 126 static unsigned long dirty_bytes_min = 2 * PAGE_SIZE;
 127
 128 /* this is needed for the proc_dointvec_minmax for [fs_]overflow UID and GID */
 129 static int maxolduid = 65535;
 130 static int minolduid;
 131 static int min_percpu_pagelist_fract = 8;
 132
 133 static int ngroups_max = NGROUPS_MAX;
 134
 135 #ifdef CONFIG_INOTIFY_USER
 136 #include <linux/inotify.h>
 137 #endif
 138 #ifdef CONFIG_SPARC
 139 #include <asm/system.h>
 140 #endif
 141
 142 #ifdef CONFIG_SPARC64
 143 extern int sysctl_tsb_ratio;
 144 #endif
 145
 146 #ifdef __hppa__
 147 extern int pwrsw_enabled;
 148 extern int unaligned_enabled;
 149 #endif
 150
 151 #ifdef CONFIG_S390
 152 #ifdef CONFIG_MATHEMU
 153 extern int sysctl_ieee_emulation_warnings;
 154 #endif
 155 extern int sysctl_userprocess_debug;
 156 extern int spin_retry;
 157 #endif
 158
 159 #ifdef CONFIG_IA64
 160 extern int no_unaligned_warning;
 161 extern int unaligned_dump_stack;
 162 #endif
 163
 164 #ifdef CONFIG_PROC_SYSCTL
 165 static int proc_do_cad_pid(struct ctl_table *table, int write,
 166                   void __user *buffer, size_t *lenp, loff_t *ppos);
 167 static int proc_taint(struct ctl_table *table, int write,
 168                                void __user *buffer, size_t *lenp, loff_t *ppos);
 169 #endif
 170
 171 #ifdef CONFIG_MAGIC_SYSRQ
 172 static int __sysrq_enabled; /* Note: sysrq code ises it's own private copy */
 173
 174 static int sysrq_sysctl_handler(ctl_table *table, int write,
 175                                 void __user *buffer, size_t *lenp,
 176                                 loff_t *ppos)
 177 {
 178         int error;
 179
 180         error = proc_dointvec(table, write, buffer, lenp, ppos);
 181         if (error)
 182                 return error;
 183
 184         if (write)
 185                 sysrq_toggle_support(__sysrq_enabled);
 186
 187         return 0;
 188 }
 189
 190 #endif
 191
 192 static struct ctl_table root_table[];
 193 static struct ctl_table_root sysctl_table_root;
 194 static struct ctl_table_header root_table_header = {
 195         .count = 1,
 196         .ctl_table = root_table,
 197         .ctl_entry = LIST_HEAD_INIT(sysctl_table_root.default_set.list),
 198         .root = &sysctl_table_root,
 199         .set = &sysctl_table_root.default_set,
 200 };
 201 static struct ctl_table_root sysctl_table_root = {
 202         .root_list = LIST_HEAD_INIT(sysctl_table_root.root_list),
 203         .default_set.list = LIST_HEAD_INIT(root_table_header.ctl_entry),
 204 };
 205
 206 static struct ctl_table kern_table[];
 207 static struct ctl_table vm_table[];
 208 static struct ctl_table fs_table[];
 209 static struct ctl_table debug_table[];
 210 static struct ctl_table dev_table[];
 211 extern struct ctl_table random_table[];
 212 #ifdef CONFIG_EPOLL
 213 extern struct ctl_table epoll_table[];
 214 #endif
 215
 216 #ifdef HAVE_ARCH_PICK_MMAP_LAYOUT
 217 int sysctl_legacy_va_layout;
 218 #endif
 219
 220 /* The default sysctl tables: */
 221
 222 static struct ctl_table root_table[] = {
 223         {
 224                 .procname       = "kernel",
 225                 .mode           = 0555,
 226                 .child          = kern_table,
 227         },
 228         {
 229                 .procname       = "vm",
 230                 .mode           = 0555,
 231                 .child          = vm_table,
 232         },
 233         {
 234                 .procname       = "fs",
 235                 .mode           = 0555,
 236                 .child          = fs_table,
 237         },
 238         {
 239                 .procname       = "debug",
 240                 .mode           = 0555,
 241                 .child          = debug_table,
 242         },
 243         {
 244                 .procname       = "dev",
 245                 .mode           = 0555,
 246                 .child          = dev_table,
 247         },
 248 /*
 249  * NOTE: do not add new entries to this table unless you have read
 250  * Documentation/sysctl/ctl_unnumbered.txt
 251  */
 252         { }
 253 };
 254
 255 #ifdef CONFIG_SCHED_DEBUG
 256 static int min_sched_granularity_ns = 100000;           /* 100 usecs */
 257 static int max_sched_granularity_ns = NSEC_PER_SEC;     /* 1 second */
 258 static int min_wakeup_granularity_ns;                   /* 0 usecs */
 259 static int max_wakeup_granularity_ns = NSEC_PER_SEC;    /* 1 second */
 260 static int min_sched_tunable_scaling = SCHED_TUNABLESCALING_NONE;
 261 static int max_sched_tunable_scaling = SCHED_TUNABLESCALING_END-1;
 262 static int min_sched_shares_ratelimit = 100000; /* 100 usec */
 263 static int max_sched_shares_ratelimit = NSEC_PER_SEC; /* 1 second */
 264 #endif
 265
 266 #ifdef CONFIG_COMPACTION
 267 static int min_extfrag_threshold;
 268 static int max_extfrag_threshold = 1000;
 269 #endif
 270
 271 static struct ctl_table kern_table[] = {
 272         {
 273                 .procname       = "sched_child_runs_first",
 274                 .data           = &sysctl_sched_child_runs_first,
 275                 .maxlen         = sizeof(unsigned int),
 276                 .mode           = 0644,
 277                 .proc_handler   = proc_dointvec,
 278         },
 279 #ifdef CONFIG_SCHED_DEBUG
 280         {
 281                 .procname       = "sched_min_granularity_ns",
 282                 .data           = &sysctl_sched_min_granularity,
 283                 .maxlen         = sizeof(unsigned int),
 284                 .mode           = 0644,
 285                 .proc_handler   = sched_proc_update_handler,
 286                 .extra1         = &min_sched_granularity_ns,
 287                 .extra2         = &max_sched_granularity_ns,
 288         },
 289         {
 290                 .procname       = "sched_latency_ns",
 291                 .data           = &sysctl_sched_latency,
 292                 .maxlen         = sizeof(unsigned int),
 293                 .mode           = 0644,
 294                 .proc_handler   = sched_proc_update_handler,
 295                 .extra1         = &min_sched_granularity_ns,
 296                 .extra2         = &max_sched_granularity_ns,
 297         },
 298         {
 299                 .procname       = "sched_wakeup_granularity_ns",
 300                 .data           = &sysctl_sched_wakeup_granularity,
 301                 .maxlen         = sizeof(unsigned int),
 302                 .mode           = 0644,
 303                 .proc_handler   = sched_proc_update_handler,
 304                 .extra1         = &min_wakeup_granularity_ns,
 305                 .extra2         = &max_wakeup_granularity_ns,
 306         },
 307         {
 308                 .procname       = "sched_shares_ratelimit",
 309                 .data           = &sysctl_sched_shares_ratelimit,
 310                 .maxlen         = sizeof(unsigned int),
 311                 .mode           = 0644,
 312                 .proc_handler   = sched_proc_update_handler,
 313                 .extra1         = &min_sched_shares_ratelimit,
 314                 .extra2         = &max_sched_shares_ratelimit,
 315         },
 316         {
 317                 .procname       = "sched_tunable_scaling",
 318                 .data           = &sysctl_sched_tunable_scaling,
 319                 .maxlen         = sizeof(enum sched_tunable_scaling),
 320                 .mode           = 0644,
 321                 .proc_handler   = sched_proc_update_handler,
 322                 .extra1         = &min_sched_tunable_scaling,
 323                 .extra2         = &max_sched_tunable_scaling,
 324         },
 325         {
 326                 .procname       = "sched_shares_thresh",
 327                 .data           = &sysctl_sched_shares_thresh,
 328                 .maxlen         = sizeof(unsigned int),
 329                 .mode           = 0644,
 330                 .proc_handler   = proc_dointvec_minmax,
 331                 .extra1         = &zero,
 332         },
 333         {
 334                 .procname       = "sched_migration_cost",
 335                 .data           = &sysctl_sched_migration_cost,
 336                 .maxlen         = sizeof(unsigned int),
 337                 .mode           = 0644,
 338                 .proc_handler   = proc_dointvec,
 339         },
 340         {
 341                 .procname       = "sched_nr_migrate",
 342                 .data           = &sysctl_sched_nr_migrate,
 343                 .maxlen         = sizeof(unsigned int),
 344                 .mode           = 0644,
 345                 .proc_handler   = proc_dointvec,
 346         },
 347         {
 348                 .procname       = "sched_time_avg",
 349                 .data           = &sysctl_sched_time_avg,
 350                 .maxlen         = sizeof(unsigned int),
 351                 .mode           = 0644,
 352                 .proc_handler   = proc_dointvec,
 353         },
 354         {
 355                 .procname       = "timer_migration",
 356                 .data           = &sysctl_timer_migration,
 357                 .maxlen         = sizeof(unsigned int),
 358                 .mode           = 0644,
 359                 .proc_handler   = proc_dointvec_minmax,
 360                 .extra1         = &zero,
 361                 .extra2         = &one,
 362         },
 363 #endif
 364         {
 365                 .procname       = "sched_rt_period_us",
 366                 .data           = &sysctl_sched_rt_period,
 367                 .maxlen         = sizeof(unsigned int),
 368                 .mode           = 0644,
 369                 .proc_handler   = sched_rt_handler,
 370         },
 371         {
 372                 .procname       = "sched_rt_runtime_us",
 373                 .data           = &sysctl_sched_rt_runtime,
 374                 .maxlen         = sizeof(int),
 375                 .mode           = 0644,
 376                 .proc_handler   = sched_rt_handler,
 377         },
 378         {
 379                 .procname       = "sched_compat_yield",
 380                 .data           = &sysctl_sched_compat_yield,
 381                 .maxlen         = sizeof(unsigned int),
 382                 .mode           = 0644,
 383                 .proc_handler   = proc_dointvec,
 384         },
 385 #ifdef CONFIG_PROVE_LOCKING
 386         {
 387                 .procname       = "prove_locking",
 388                 .data           = &prove_locking,
 389                 .maxlen         = sizeof(int),
 390                 .mode           = 0644,
 391                 .proc_handler   = proc_dointvec,
 392         },
 393 #endif
 394 #ifdef CONFIG_LOCK_STAT
 395         {
 396                 .procname       = "lock_stat",
 397                 .data           = &lock_stat,
 398                 .maxlen         = sizeof(int),
 399                 .mode           = 0644,
 400                 .proc_handler   = proc_dointvec,
 401         },
 402 #endif
 403         {
 404                 .procname       = "panic",
 405                 .data           = &panic_timeout,
 406                 .maxlen         = sizeof(int),
 407                 .mode           = 0644,
 408                 .proc_handler   = proc_dointvec,
 409         },
 410         {
 411                 .procname       = "core_uses_pid",
 412                 .data           = &core_uses_pid,
 413                 .maxlen         = sizeof(int),
 414                 .mode           = 0644,
 415                 .proc_handler   = proc_dointvec,
 416         },
 417         {
 418                 .procname       = "core_pattern",
 419                 .data           = core_pattern,
 420                 .maxlen         = CORENAME_MAX_SIZE,
 421                 .mode           = 0644,
 422                 .proc_handler   = proc_dostring,
 423         },
 424         {
 425                 .procname       = "core_pipe_limit",
 426                 .data           = &core_pipe_limit,
 427                 .maxlen         = sizeof(unsigned int),
 428                 .mode           = 0644,
 429                 .proc_handler   = proc_dointvec,
 430         },
 431 #ifdef CONFIG_PROC_SYSCTL
 432         {
 433                 .procname       = "tainted",
 434                 .maxlen         = sizeof(long),
 435                 .mode           = 0644,
 436                 .proc_handler   = proc_taint,
 437         },
 438 #endif
 439 #ifdef CONFIG_LATENCYTOP
 440         {
 441                 .procname       = "latencytop",
 442                 .data           = &latencytop_enabled,
 443                 .maxlen         = sizeof(int),
 444                 .mode           = 0644,
 445                 .proc_handler   = proc_dointvec,
 446         },
 447 #endif
 448 #ifdef CONFIG_BLK_DEV_INITRD
 449         {
 450                 .procname       = "real-root-dev",
 451                 .data           = &real_root_dev,
 452                 .maxlen         = sizeof(int),
 453                 .mode           = 0644,
 454                 .proc_handler   = proc_dointvec,
 455         },
 456 #endif
 457         {
 458                 .procname       = "print-fatal-signals",
 459                 .data           = &print_fatal_signals,
 460                 .maxlen         = sizeof(int),
 461                 .mode           = 0644,
 462                 .proc_handler   = proc_dointvec,
 463         },
 464 #ifdef CONFIG_SPARC
 465         {
 466                 .procname       = "reboot-cmd",
 467                 .data           = reboot_command,
 468                 .maxlen         = 256,
 469                 .mode           = 0644,
 470                 .proc_handler   = proc_dostring,
 471         },
 472         {
 473                 .procname       = "stop-a",
 474                 .data           = &stop_a_enabled,
 475                 .maxlen         = sizeof (int),
 476                 .mode           = 0644,
 477                 .proc_handler   = proc_dointvec,
 478         },
 479         {
 480                 .procname       = "scons-poweroff",
 481                 .data           = &scons_pwroff,
 482                 .maxlen         = sizeof (int),
 483                 .mode           = 0644,
 484                 .proc_handler   = proc_dointvec,
 485         },
 486 #endif
 487 #ifdef CONFIG_SPARC64
 488         {
 489                 .procname       = "tsb-ratio",
 490                 .data           = &sysctl_tsb_ratio,
 491                 .maxlen         = sizeof (int),
 492                 .mode           = 0644,
 493                 .proc_handler   = proc_dointvec,
 494         },
 495 #endif
 496 #ifdef __hppa__
 497         {
 498                 .procname       = "soft-power",
 499                 .data           = &pwrsw_enabled,
 500                 .maxlen         = sizeof (int),
 501                 .mode           = 0644,
 502                 .proc_handler   = proc_dointvec,
 503         },
 504         {
 505                 .procname       = "unaligned-trap",
 506                 .data           = &unaligned_enabled,
 507                 .maxlen         = sizeof (int),
 508                 .mode           = 0644,
 509                 .proc_handler   = proc_dointvec,
 510         },
 511 #endif
 512         {
 513                 .procname       = "ctrl-alt-del",
 514                 .data           = &C_A_D,
 515                 .maxlen         = sizeof(int),
 516                 .mode           = 0644,
 517                 .proc_handler   = proc_dointvec,
 518         },
 519 #ifdef CONFIG_FUNCTION_TRACER
 520         {
 521                 .procname       = "ftrace_enabled",
 522                 .data           = &ftrace_enabled,
 523                 .maxlen         = sizeof(int),
 524                 .mode           = 0644,
 525                 .proc_handler   = ftrace_enable_sysctl,
 526         },
 527 #endif
 528 #ifdef CONFIG_STACK_TRACER
 529         {
 530                 .procname       = "stack_tracer_enabled",
 531                 .data           = &stack_tracer_enabled,
 532                 .maxlen         = sizeof(int),
 533                 .mode           = 0644,
 534                 .proc_handler   = stack_trace_sysctl,
 535         },
 536 #endif
 537 #ifdef CONFIG_TRACING
 538         {
 539                 .procname       = "ftrace_dump_on_oops",
 540                 .data           = &ftrace_dump_on_oops,
 541                 .maxlen         = sizeof(int),
 542                 .mode           = 0644,
 543                 .proc_handler   = proc_dointvec,
 544         },
 545 #endif
 546 #ifdef CONFIG_MODULES
 547         {
 548                 .procname       = "modprobe",
 549                 .data           = &modprobe_path,
 550                 .maxlen         = KMOD_PATH_LEN,
 551                 .mode           = 0644,
 552                 .proc_handler   = proc_dostring,
 553         },
 554         {
 555                 .procname       = "modules_disabled",
 556                 .data           = &modules_disabled,
 557                 .maxlen         = sizeof(int),
 558                 .mode           = 0644,
 559                 /* only handle a transition from default "0" to "1" */
 560                 .proc_handler   = proc_dointvec_minmax,
 561                 .extra1         = &one,
 562                 .extra2         = &one,
 563         },
 564 #endif
 565 #ifdef CONFIG_HOTPLUG
 566         {
 567                 .procname       = "hotplug",
 568                 .data           = &uevent_helper,
 569                 .maxlen         = UEVENT_HELPER_PATH_LEN,
 570                 .mode           = 0644,
 571                 .proc_handler   = proc_dostring,
 572         },
 573 #endif
 574 #ifdef CONFIG_CHR_DEV_SG
 575         {
 576                 .procname       = "sg-big-buff",
 577                 .data           = &sg_big_buff,
 578                 .maxlen         = sizeof (int),
 579                 .mode           = 0444,
 580                 .proc_handler   = proc_dointvec,
 581         },
 582 #endif
 583 #ifdef CONFIG_BSD_PROCESS_ACCT
 584         {
 585                 .procname       = "acct",
 586                 .data           = &acct_parm,
 587                 .maxlen         = 3*sizeof(int),
 588                 .mode           = 0644,
 589                 .proc_handler   = proc_dointvec,
 590         },
 591 #endif
 592 #ifdef CONFIG_MAGIC_SYSRQ
 593         {
 594                 .procname       = "sysrq",
 595                 .data           = &__sysrq_enabled,
 596                 .maxlen         = sizeof (int),
 597                 .mode           = 0644,
 598                 .proc_handler   = sysrq_sysctl_handler,
 599         },
 600 #endif
 601 #ifdef CONFIG_PROC_SYSCTL
 602         {
 603                 .procname       = "cad_pid",
 604                 .data           = NULL,
 605                 .maxlen         = sizeof (int),
 606                 .mode           = 0600,
 607                 .proc_handler   = proc_do_cad_pid,
 608         },
 609 #endif
 610         {
 611                 .procname       = "threads-max",
 612                 .data           = &max_threads,
 613                 .maxlen         = sizeof(int),
 614                 .mode           = 0644,
 615                 .proc_handler   = proc_dointvec,
 616         },
 617         {
 618                 .procname       = "random",
 619                 .mode           = 0555,
 620                 .child          = random_table,
 621         },
 622         {
 623                 .procname       = "overflowuid",
 624                 .data           = &overflowuid,
 625                 .maxlen         = sizeof(int),
 626                 .mode           = 0644,
 627                 .proc_handler   = proc_dointvec_minmax,
 628                 .extra1         = &minolduid,
 629                 .extra2         = &maxolduid,
 630         },
 631         {
 632                 .procname       = "overflowgid",
 633                 .data           = &overflowgid,
 634                 .maxlen         = sizeof(int),
 635                 .mode           = 0644,
 636                 .proc_handler   = proc_dointvec_minmax,
 637                 .extra1         = &minolduid,
 638                 .extra2         = &maxolduid,
 639         },
 640 #ifdef CONFIG_S390
 641 #ifdef CONFIG_MATHEMU
 642         {
 643                 .procname       = "ieee_emulation_warnings",
 644                 .data           = &sysctl_ieee_emulation_warnings,
 645                 .maxlen         = sizeof(int),
 646                 .mode           = 0644,
 647                 .proc_handler   = proc_dointvec,
 648         },
 649 #endif
 650         {
 651                 .procname       = "userprocess_debug",
 652                 .data           = &show_unhandled_signals,
 653                 .maxlen         = sizeof(int),
 654                 .mode           = 0644,
 655                 .proc_handler   = proc_dointvec,
 656         },
 657 #endif
 658         {
 659                 .procname       = "pid_max",
 660                 .data           = &pid_max,
 661                 .maxlen         = sizeof (int),
 662                 .mode           = 0644,
 663                 .proc_handler   = proc_dointvec_minmax,
 664                 .extra1         = &pid_max_min,
 665                 .extra2         = &pid_max_max,
 666         },
 667         {
 668                 .procname       = "panic_on_oops",
 669                 .data           = &panic_on_oops,
 670                 .maxlen         = sizeof(int),
 671                 .mode           = 0644,
 672                 .proc_handler   = proc_dointvec,
 673         },
 674 #if defined CONFIG_PRINTK
 675         {
 676                 .procname       = "printk",
 677                 .data           = &console_loglevel,
 678                 .maxlen         = 4*sizeof(int),
 679                 .mode           = 0644,
 680                 .proc_handler   = proc_dointvec,
 681         },
 682         {
 683                 .procname       = "printk_ratelimit",
 684                 .data           = &printk_ratelimit_state.interval,
 685                 .maxlen         = sizeof(int),
 686                 .mode           = 0644,
 687                 .proc_handler   = proc_dointvec_jiffies,
 688         },
 689         {
 690                 .procname       = "printk_ratelimit_burst",
 691                 .data           = &printk_ratelimit_state.burst,
 692                 .maxlen         = sizeof(int),
 693                 .mode           = 0644,
 694                 .proc_handler   = proc_dointvec,
 695         },
 696         {
 697                 .procname       = "printk_delay",
 698                 .data           = &printk_delay_msec,
 699                 .maxlen         = sizeof(int),
 700                 .mode           = 0644,
 701                 .proc_handler   = proc_dointvec_minmax,
 702                 .extra1         = &zero,
 703                 .extra2         = &ten_thousand,
 704         },
 705         {
 706                 .procname       = "dmesg_restrict",
 707                 .data           = &dmesg_restrict,
 708                 .maxlen         = sizeof(int),
 709                 .mode           = 0644,
 710                 .proc_handler   = proc_dointvec_minmax,
 711                 .extra1         = &zero,
 712                 .extra2         = &one,
 713         },
 714 #endif
 715         {
 716                 .procname       = "ngroups_max",
 717                 .data           = &ngroups_max,
 718                 .maxlen         = sizeof (int),
 719                 .mode           = 0444,
 720                 .proc_handler   = proc_dointvec,
 721         },
 722 #if defined(CONFIG_LOCKUP_DETECTOR)
 723         {
 724                 .procname       = "watchdog",
 725                 .data           = &watchdog_enabled,
 726                 .maxlen         = sizeof (int),
 727                 .mode           = 0644,
 728                 .proc_handler   = proc_dowatchdog_enabled,
 729         },
 730         {
 731                 .procname       = "watchdog_thresh",
 732                 .data           = &softlockup_thresh,
 733                 .maxlen         = sizeof(int),
 734                 .mode           = 0644,
 735                 .proc_handler   = proc_dowatchdog_thresh,
 736                 .extra1         = &neg_one,
 737                 .extra2         = &sixty,
 738         },
 739         {
 740                 .procname       = "softlockup_panic",
 741                 .data           = &softlockup_panic,
 742                 .maxlen         = sizeof(int),
 743                 .mode           = 0644,
 744                 .proc_handler   = proc_dointvec_minmax,
 745                 .extra1         = &zero,
 746                 .extra2         = &one,
 747         },
 748 #endif
 749 #if defined(CONFIG_X86_LOCAL_APIC) && defined(CONFIG_X86) && !defined(CONFIG_LOCKUP_DETECTOR)
 750         {
 751                 .procname       = "unknown_nmi_panic",
 752                 .data           = &unknown_nmi_panic,
 753                 .maxlen         = sizeof (int),
 754                 .mode           = 0644,
 755                 .proc_handler   = proc_dointvec,
 756         },
 757         {
 758                 .procname       = "nmi_watchdog",
 759                 .data           = &nmi_watchdog_enabled,
 760                 .maxlen         = sizeof (int),
 761                 .mode           = 0644,
 762                 .proc_handler   = proc_nmi_enabled,
 763         },
 764 #endif
 765 #if defined(CONFIG_X86)
 766         {
 767                 .procname       = "panic_on_unrecovered_nmi",
 768                 .data           = &panic_on_unrecovered_nmi,
 769                 .maxlen         = sizeof(int),
 770                 .mode           = 0644,
 771                 .proc_handler   = proc_dointvec,
 772         },
 773         {
 774                 .procname       = "panic_on_io_nmi",
 775                 .data           = &panic_on_io_nmi,
 776                 .maxlen         = sizeof(int),
 777                 .mode           = 0644,
 778                 .proc_handler   = proc_dointvec,
 779         },
 780         {
 781                 .procname       = "bootloader_type",
 782                 .data           = &bootloader_type,
 783                 .maxlen         = sizeof (int),
 784                 .mode           = 0444,
 785                 .proc_handler   = proc_dointvec,
 786         },
 787         {
 788                 .procname       = "bootloader_version",
 789                 .data           = &bootloader_version,
 790                 .maxlen         = sizeof (int),
 791                 .mode           = 0444,
 792                 .proc_handler   = proc_dointvec,
 793         },
 794         {
 795                 .procname       = "kstack_depth_to_print",
 796                 .data           = &kstack_depth_to_print,
 797                 .maxlen         = sizeof(int),
 798                 .mode           = 0644,
 799                 .proc_handler   = proc_dointvec,
 800         },
 801         {
 802                 .procname       = "io_delay_type",
 803                 .data           = &io_delay_type,
 804                 .maxlen         = sizeof(int),
 805                 .mode           = 0644,
 806                 .proc_handler   = proc_dointvec,
 807         },
 808 #endif
 809 #if defined(CONFIG_MMU)
 810         {
 811                 .procname       = "randomize_va_space",
 812                 .data           = &randomize_va_space,
 813                 .maxlen         = sizeof(int),
 814                 .mode           = 0644,
 815                 .proc_handler   = proc_dointvec,
 816         },
 817 #endif
 818 #if defined(CONFIG_S390) && defined(CONFIG_SMP)
 819         {
 820                 .procname       = "spin_retry",
 821                 .data           = &spin_retry,
 822                 .maxlen         = sizeof (int),
 823                 .mode           = 0644,
 824                 .proc_handler   = proc_dointvec,
 825         },
 826 #endif
 827 #if     defined(CONFIG_ACPI_SLEEP) && defined(CONFIG_X86)
 828         {
 829                 .procname       = "acpi_video_flags",
 830                 .data           = &acpi_realmode_flags,
 831                 .maxlen         = sizeof (unsigned long),
 832                 .mode           = 0644,
 833                 .proc_handler   = proc_doulongvec_minmax,
 834         },
 835 #endif
 836 #ifdef CONFIG_IA64
 837         {
 838                 .procname       = "ignore-unaligned-usertrap",
 839                 .data           = &no_unaligned_warning,
 840                 .maxlen         = sizeof (int),
 841                 .mode           = 0644,
 842                 .proc_handler   = proc_dointvec,
 843         },
 844         {
 845                 .procname       = "unaligned-dump-stack",
 846                 .data           = &unaligned_dump_stack,
 847                 .maxlen         = sizeof (int),
 848                 .mode           = 0644,
 849                 .proc_handler   = proc_dointvec,
 850         },
 851 #endif
 852 #ifdef CONFIG_DETECT_HUNG_TASK
 853         {
 854                 .procname       = "hung_task_panic",
 855                 .data           = &sysctl_hung_task_panic,
 856                 .maxlen         = sizeof(int),
 857                 .mode           = 0644,
 858                 .proc_handler   = proc_dointvec_minmax,
 859                 .extra1         = &zero,
 860                 .extra2         = &one,
 861         },
 862         {
 863                 .procname       = "hung_task_check_count",
 864                 .data           = &sysctl_hung_task_check_count,
 865                 .maxlen         = sizeof(unsigned long),
 866                 .mode           = 0644,
 867                 .proc_handler   = proc_doulongvec_minmax,
 868         },
 869         {
 870                 .procname       = "hung_task_timeout_secs",
 871                 .data           = &sysctl_hung_task_timeout_secs,
 872                 .maxlen         = sizeof(unsigned long),
 873                 .mode           = 0644,
 874                 .proc_handler   = proc_dohung_task_timeout_secs,
 875         },
 876         {
 877                 .procname       = "hung_task_warnings",
 878                 .data           = &sysctl_hung_task_warnings,
 879                 .maxlen         = sizeof(unsigned long),
 880                 .mode           = 0644,
 881                 .proc_handler   = proc_doulongvec_minmax,
 882         },
 883 #endif
 884 #ifdef CONFIG_COMPAT
 885         {
 886                 .procname       = "compat-log",
 887                 .data           = &compat_log,
 888                 .maxlen         = sizeof (int),
 889                 .mode           = 0644,
 890                 .proc_handler   = proc_dointvec,
 891         },
 892 #endif
 893 #ifdef CONFIG_RT_MUTEXES
 894         {
 895                 .procname       = "max_lock_depth",
 896                 .data           = &max_lock_depth,
 897                 .maxlen         = sizeof(int),
 898                 .mode           = 0644,
 899                 .proc_handler   = proc_dointvec,
 900         },
 901 #endif
 902         {
 903                 .procname       = "poweroff_cmd",
 904                 .data           = &poweroff_cmd,
 905                 .maxlen         = POWEROFF_CMD_PATH_LEN,
 906                 .mode           = 0644,
 907                 .proc_handler   = proc_dostring,
 908         },
 909 #ifdef CONFIG_KEYS
 910         {
 911                 .procname       = "keys",
 912                 .mode           = 0555,
 913                 .child          = key_sysctls,
 914         },
 915 #endif
 916 #ifdef CONFIG_RCU_TORTURE_TEST
 917         {
 918                 .procname       = "rcutorture_runnable",
 919                 .data           = &rcutorture_runnable,
 920                 .maxlen         = sizeof(int),
 921                 .mode           = 0644,
 922                 .proc_handler   = proc_dointvec,
 923         },
 924 #endif
 925 #ifdef CONFIG_PERF_EVENTS
 926         {
 927                 .procname       = "perf_event_paranoid",
 928                 .data           = &sysctl_perf_event_paranoid,
 929                 .maxlen         = sizeof(sysctl_perf_event_paranoid),
 930                 .mode           = 0644,
 931                 .proc_handler   = proc_dointvec,
 932         },
 933         {
 934                 .procname       = "perf_event_mlock_kb",
 935                 .data           = &sysctl_perf_event_mlock,
 936                 .maxlen         = sizeof(sysctl_perf_event_mlock),
 937                 .mode           = 0644,
 938                 .proc_handler   = proc_dointvec,
 939         },
 940         {
 941                 .procname       = "perf_event_max_sample_rate",
 942                 .data           = &sysctl_perf_event_sample_rate,
 943                 .maxlen         = sizeof(sysctl_perf_event_sample_rate),
 944                 .mode           = 0644,
 945                 .proc_handler   = proc_dointvec,
 946         },
 947 #endif
 948 #ifdef CONFIG_KMEMCHECK
 949         {
 950                 .procname       = "kmemcheck",
 951                 .data           = &kmemcheck_enabled,
 952                 .maxlen         = sizeof(int),
 953                 .mode           = 0644,
 954                 .proc_handler   = proc_dointvec,
 955         },
 956 #endif
 957 #ifdef CONFIG_BLOCK
 958         {
 959                 .procname       = "blk_iopoll",
 960                 .data           = &blk_iopoll_enabled,
 961                 .maxlen         = sizeof(int),
 962                 .mode           = 0644,
 963                 .proc_handler   = proc_dointvec,
 964         },
 965 #endif
 966 /*
 967  * NOTE: do not add new entries to this table unless you have read
 968  * Documentation/sysctl/ctl_unnumbered.txt
 969  */
 970         { }
 971 };
 972
 973 static struct ctl_table vm_table[] = {
 974         {
 975                 .procname       = "overcommit_memory",
 976                 .data           = &sysctl_overcommit_memory,
 977                 .maxlen         = sizeof(sysctl_overcommit_memory),
 978                 .mode           = 0644,
 979                 .proc_handler   = proc_dointvec,
 980         },
 981         {
 982                 .procname       = "panic_on_oom",
 983                 .data           = &sysctl_panic_on_oom,
 984                 .maxlen         = sizeof(sysctl_panic_on_oom),
 985                 .mode           = 0644,
 986                 .proc_handler   = proc_dointvec,
 987         },
 988         {
 989                 .procname       = "oom_kill_allocating_task",
 990                 .data           = &sysctl_oom_kill_allocating_task,
 991                 .maxlen         = sizeof(sysctl_oom_kill_allocating_task),
 992                 .mode           = 0644,
 993                 .proc_handler   = proc_dointvec,
 994         },
 995         {
 996                 .procname       = "oom_dump_tasks",
 997                 .data           = &sysctl_oom_dump_tasks,
 998                 .maxlen         = sizeof(sysctl_oom_dump_tasks),
 999                 .mode           = 0644,
1000                 .proc_handler   = proc_dointvec,
1001         },
1002         {
1003                 .procname       = "overcommit_ratio",
1004                 .data           = &sysctl_overcommit_ratio,
1005                 .maxlen         = sizeof(sysctl_overcommit_ratio),
1006                 .mode           = 0644,
1007                 .proc_handler   = proc_dointvec,
1008         },
1009         {
1010                 .procname       = "page-cluster",
1011                 .data           = &page_cluster,
1012                 .maxlen         = sizeof(int),
1013                 .mode           = 0644,
1014                 .proc_handler   = proc_dointvec,
1015         },
1016         {
1017                 .procname       = "dirty_background_ratio",
1018                 .data           = &dirty_background_ratio,
1019                 .maxlen         = sizeof(dirty_background_ratio),
1020                 .mode           = 0644,
1021                 .proc_handler   = dirty_background_ratio_handler,
1022                 .extra1         = &zero,
1023                 .extra2         = &one_hundred,
1024         },
1025         {
1026                 .procname       = "dirty_background_bytes",
1027                 .data           = &dirty_background_bytes,
1028                 .maxlen         = sizeof(dirty_background_bytes),
1029                 .mode           = 0644,
1030                 .proc_handler   = dirty_background_bytes_handler,
1031                 .extra1         = &one_ul,
1032         },
1033         {
1034                 .procname       = "dirty_ratio",
1035                 .data           = &vm_dirty_ratio,
1036                 .maxlen         = sizeof(vm_dirty_ratio),
1037                 .mode           = 0644,
1038                 .proc_handler   = dirty_ratio_handler,
1039                 .extra1         = &zero,
1040                 .extra2         = &one_hundred,
1041         },
1042         {
1043                 .procname       = "dirty_bytes",
1044                 .data           = &vm_dirty_bytes,
1045                 .maxlen         = sizeof(vm_dirty_bytes),
1046                 .mode           = 0644,
1047                 .proc_handler   = dirty_bytes_handler,
1048                 .extra1         = &dirty_bytes_min,
1049         },
1050         {
1051                 .procname       = "dirty_writeback_centisecs",
1052                 .data           = &dirty_writeback_interval,
1053                 .maxlen         = sizeof(dirty_writeback_interval),
1054                 .mode           = 0644,
1055                 .proc_handler   = dirty_writeback_centisecs_handler,
1056         },
1057         {
1058                 .procname       = "dirty_expire_centisecs",
1059                 .data           = &dirty_expire_interval,
1060                 .maxlen         = sizeof(dirty_expire_interval),
1061                 .mode           = 0644,
1062                 .proc_handler   = proc_dointvec,
1063         },
1064         {
1065                 .procname       = "nr_pdflush_threads",
1066                 .data           = &nr_pdflush_threads,
1067                 .maxlen         = sizeof nr_pdflush_threads,
1068                 .mode           = 0444 /* read-only*/,
1069                 .proc_handler   = proc_dointvec,
1070         },
1071         {
1072                 .procname       = "swappiness",
1073                 .data           = &vm_swappiness,
1074                 .maxlen         = sizeof(vm_swappiness),
1075                 .mode           = 0644,
1076                 .proc_handler   = proc_dointvec_minmax,
1077                 .extra1         = &zero,
1078                 .extra2         = &one_hundred,
1079         },
1080 #ifdef CONFIG_HUGETLB_PAGE
1081         {
1082                 .procname       = "nr_hugepages",
1083                 .data           = NULL,
1084                 .maxlen         = sizeof(unsigned long),
1085                 .mode           = 0644,
1086                 .proc_handler   = hugetlb_sysctl_handler,
1087                 .extra1         = (void *)&hugetlb_zero,
1088                 .extra2         = (void *)&hugetlb_infinity,
1089         },
1090 #ifdef CONFIG_NUMA
1091         {
1092                 .procname       = "nr_hugepages_mempolicy",
1093                 .data           = NULL,
1094                 .maxlen         = sizeof(unsigned long),
1095                 .mode           = 0644,
1096                 .proc_handler   = &hugetlb_mempolicy_sysctl_handler,
1097                 .extra1         = (void *)&hugetlb_zero,
1098                 .extra2         = (void *)&hugetlb_infinity,
1099         },
1100 #endif
1101          {
1102                 .procname       = "hugetlb_shm_group",
1103                 .data           = &sysctl_hugetlb_shm_group,
1104                 .maxlen         = sizeof(gid_t),
1105                 .mode           = 0644,
1106                 .proc_handler   = proc_dointvec,
1107          },
1108          {
1109                 .procname       = "hugepages_treat_as_movable",
1110                 .data           = &hugepages_treat_as_movable,
1111                 .maxlen         = sizeof(int),
1112                 .mode           = 0644,
1113                 .proc_handler   = hugetlb_treat_movable_handler,
1114         },
1115         {
1116                 .procname       = "nr_overcommit_hugepages",
1117                 .data           = NULL,
1118                 .maxlen         = sizeof(unsigned long),
1119                 .mode           = 0644,
1120                 .proc_handler   = hugetlb_overcommit_handler,
1121                 .extra1         = (void *)&hugetlb_zero,
1122                 .extra2         = (void *)&hugetlb_infinity,
1123         },
1124 #endif
1125         {
1126                 .procname       = "lowmem_reserve_ratio",
1127                 .data           = &sysctl_lowmem_reserve_ratio,
1128                 .maxlen         = sizeof(sysctl_lowmem_reserve_ratio),
1129                 .mode           = 0644,
1130                 .proc_handler   = lowmem_reserve_ratio_sysctl_handler,
1131         },
1132         {
1133                 .procname       = "drop_caches",
1134                 .data           = &sysctl_drop_caches,
1135                 .maxlen         = sizeof(int),
1136                 .mode           = 0644,
1137                 .proc_handler   = drop_caches_sysctl_handler,
1138         },
1139 #ifdef CONFIG_COMPACTION
1140         {
1141                 .procname       = "compact_memory",
1142                 .data           = &sysctl_compact_memory,
1143                 .maxlen         = sizeof(int),
1144                 .mode           = 0200,
1145                 .proc_handler   = sysctl_compaction_handler,
1146         },
1147         {
1148                 .procname       = "extfrag_threshold",
1149                 .data           = &sysctl_extfrag_threshold,
1150                 .maxlen         = sizeof(int),
1151                 .mode           = 0644,
1152                 .proc_handler   = sysctl_extfrag_handler,
1153                 .extra1         = &min_extfrag_threshold,
1154                 .extra2         = &max_extfrag_threshold,
1155         },
1156
1157 #endif /* CONFIG_COMPACTION */
1158         {
1159                 .procname       = "min_free_kbytes",
1160                 .data           = &min_free_kbytes,
1161                 .maxlen         = sizeof(min_free_kbytes),
1162                 .mode           = 0644,
1163                 .proc_handler   = min_free_kbytes_sysctl_handler,
1164                 .extra1         = &zero,
1165         },
1166         {
1167                 .procname       = "percpu_pagelist_fraction",
1168                 .data           = &percpu_pagelist_fraction,
1169                 .maxlen         = sizeof(percpu_pagelist_fraction),
1170                 .mode           = 0644,
1171                 .proc_handler   = percpu_pagelist_fraction_sysctl_handler,
1172                 .extra1         = &min_percpu_pagelist_fract,
1173         },
1174 #ifdef CONFIG_MMU
1175         {
1176                 .procname       = "max_map_count",
1177                 .data           = &sysctl_max_map_count,
1178                 .maxlen         = sizeof(sysctl_max_map_count),
1179                 .mode           = 0644,
1180                 .proc_handler   = proc_dointvec_minmax,
1181                 .extra1         = &zero,
1182         },
1183 #else
1184         {
1185                 .procname       = "nr_trim_pages",
1186                 .data           = &sysctl_nr_trim_pages,
1187                 .maxlen         = sizeof(sysctl_nr_trim_pages),
1188                 .mode           = 0644,
1189                 .proc_handler   = proc_dointvec_minmax,
1190                 .extra1         = &zero,
1191         },
1192 #endif
1193         {
1194                 .procname       = "laptop_mode",
1195                 .data           = &laptop_mode,
1196                 .maxlen         = sizeof(laptop_mode),
1197                 .mode           = 0644,
1198                 .proc_handler   = proc_dointvec_jiffies,
1199         },
1200         {
1201                 .procname       = "block_dump",
1202                 .data           = &block_dump,
1203                 .maxlen         = sizeof(block_dump),
1204                 .mode           = 0644,
1205                 .proc_handler   = proc_dointvec,
1206                 .extra1         = &zero,
1207         },
1208         {
1209                 .procname       = "vfs_cache_pressure",
1210                 .data           = &sysctl_vfs_cache_pressure,
1211                 .maxlen         = sizeof(sysctl_vfs_cache_pressure),
1212                 .mode           = 0644,
1213                 .proc_handler   = proc_dointvec,
1214                 .extra1         = &zero,
1215         },
1216 #ifdef HAVE_ARCH_PICK_MMAP_LAYOUT
1217         {
1218                 .procname       = "legacy_va_layout",
1219                 .data           = &sysctl_legacy_va_layout,
1220                 .maxlen         = sizeof(sysctl_legacy_va_layout),
1221                 .mode           = 0644,
1222                 .proc_handler   = proc_dointvec,
1223                 .extra1         = &zero,
1224         },
1225 #endif
1226 #ifdef CONFIG_NUMA
1227         {
1228                 .procname       = "zone_reclaim_mode",
1229                 .data           = &zone_reclaim_mode,
1230                 .maxlen         = sizeof(zone_reclaim_mode),
1231                 .mode           = 0644,
1232                 .proc_handler   = proc_dointvec,
1233                 .extra1         = &zero,
1234         },
1235         {
1236                 .procname       = "min_unmapped_ratio",
1237                 .data           = &sysctl_min_unmapped_ratio,
1238                 .maxlen         = sizeof(sysctl_min_unmapped_ratio),
1239                 .mode           = 0644,
1240                 .proc_handler   = sysctl_min_unmapped_ratio_sysctl_handler,
1241                 .extra1         = &zero,
1242                 .extra2         = &one_hundred,
1243         },
1244         {
1245                 .procname       = "min_slab_ratio",
1246                 .data           = &sysctl_min_slab_ratio,
1247                 .maxlen         = sizeof(sysctl_min_slab_ratio),
1248                 .mode           = 0644,
1249                 .proc_handler   = sysctl_min_slab_ratio_sysctl_handler,
1250                 .extra1         = &zero,
1251                 .extra2         = &one_hundred,
1252         },
1253 #endif
1254 #ifdef CONFIG_SMP
1255         {
1256                 .procname       = "stat_interval",
1257                 .data           = &sysctl_stat_interval,
1258                 .maxlen         = sizeof(sysctl_stat_interval),
1259                 .mode           = 0644,
1260                 .proc_handler   = proc_dointvec_jiffies,
1261         },
1262 #endif
1263 #ifdef CONFIG_MMU
1264         {
1265                 .procname       = "mmap_min_addr",
1266                 .data           = &dac_mmap_min_addr,
1267                 .maxlen         = sizeof(unsigned long),
1268                 .mode           = 0644,
1269                 .proc_handler   = mmap_min_addr_handler,
1270         },
1271 #endif
1272 #ifdef CONFIG_NUMA
1273         {
1274                 .procname       = "numa_zonelist_order",
1275                 .data           = &numa_zonelist_order,
1276                 .maxlen         = NUMA_ZONELIST_ORDER_LEN,
1277                 .mode           = 0644,
1278                 .proc_handler   = numa_zonelist_order_handler,
1279         },
1280 #endif
1281 #if (defined(CONFIG_X86_32) && !defined(CONFIG_UML))|| \
1282    (defined(CONFIG_SUPERH) && defined(CONFIG_VSYSCALL))
1283         {
1284                 .procname       = "vdso_enabled",
1285                 .data           = &vdso_enabled,
1286                 .maxlen         = sizeof(vdso_enabled),
1287                 .mode           = 0644,
1288                 .proc_handler   = proc_dointvec,
1289                 .extra1         = &zero,
1290         },
1291 #endif
1292 #ifdef CONFIG_HIGHMEM
1293         {
1294                 .procname       = "highmem_is_dirtyable",
1295                 .data           = &vm_highmem_is_dirtyable,
1296                 .maxlen         = sizeof(vm_highmem_is_dirtyable),
1297                 .mode           = 0644,
1298                 .proc_handler   = proc_dointvec_minmax,
1299                 .extra1         = &zero,
1300                 .extra2         = &one,
1301         },
1302 #endif
1303         {
1304                 .procname       = "scan_unevictable_pages",
1305                 .data           = &scan_unevictable_pages,
1306                 .maxlen         = sizeof(scan_unevictable_pages),
1307                 .mode           = 0644,
1308                 .proc_handler   = scan_unevictable_handler,
1309         },
1310 #ifdef CONFIG_MEMORY_FAILURE
1311         {
1312                 .procname       = "memory_failure_early_kill",
1313                 .data           = &sysctl_memory_failure_early_kill,
1314                 .maxlen         = sizeof(sysctl_memory_failure_early_kill),
1315                 .mode           = 0644,
1316                 .proc_handler   = proc_dointvec_minmax,
1317                 .extra1         = &zero,
1318                 .extra2         = &one,
1319         },
1320         {
1321                 .procname       = "memory_failure_recovery",
1322                 .data           = &sysctl_memory_failure_recovery,
1323                 .maxlen         = sizeof(sysctl_memory_failure_recovery),
1324                 .mode           = 0644,
1325                 .proc_handler   = proc_dointvec_minmax,
1326                 .extra1         = &zero,
1327                 .extra2         = &one,
1328         },
1329 #endif
1330
1331 /*
1332  * NOTE: do not add new entries to this table unless you have read
1333  * Documentation/sysctl/ctl_unnumbered.txt
1334  */
1335         { }
1336 };
1337
1338 #if defined(CONFIG_BINFMT_MISC) || defined(CONFIG_BINFMT_MISC_MODULE)
1339 static struct ctl_table binfmt_misc_table[] = {
1340         { }
1341 };
1342 #endif
1343
1344 static struct ctl_table fs_table[] = {
1345         {
1346                 .procname       = "inode-nr",
1347                 .data           = &inodes_stat,
1348                 .maxlen         = 2*sizeof(int),
1349                 .mode           = 0444,
1350                 .proc_handler   = proc_nr_inodes,
1351         },
1352         {
1353                 .procname       = "inode-state",
1354                 .data           = &inodes_stat,
1355                 .maxlen         = 7*sizeof(int),
1356                 .mode           = 0444,
1357                 .proc_handler   = proc_nr_inodes,
1358         },
1359         {
1360                 .procname       = "file-nr",
1361                 .data           = &files_stat,
1362                 .maxlen         = sizeof(files_stat),
1363                 .mode           = 0444,
1364                 .proc_handler   = proc_nr_files,
1365         },
1366         {
1367                 .procname       = "file-max",
1368                 .data           = &files_stat.max_files,
1369                 .maxlen         = sizeof(files_stat.max_files),
1370                 .mode           = 0644,
1371                 .proc_handler   = proc_doulongvec_minmax,
1372         },
1373         {
1374                 .procname       = "nr_open",
1375                 .data           = &sysctl_nr_open,
1376                 .maxlen         = sizeof(int),
1377                 .mode           = 0644,
1378                 .proc_handler   = proc_dointvec_minmax,
1379                 .extra1         = &sysctl_nr_open_min,
1380                 .extra2         = &sysctl_nr_open_max,
1381         },
1382         {
1383                 .procname       = "dentry-state",
1384                 .data           = &dentry_stat,
1385                 .maxlen         = 6*sizeof(int),
1386                 .mode           = 0444,
1387                 .proc_handler   = proc_nr_dentry,
1388         },
1389         {
1390                 .procname       = "overflowuid",
1391                 .data           = &fs_overflowuid,
1392                 .maxlen         = sizeof(int),
1393                 .mode           = 0644,
1394                 .proc_handler   = proc_dointvec_minmax,
1395                 .extra1         = &minolduid,
1396                 .extra2         = &maxolduid,
1397         },
1398         {
1399                 .procname       = "overflowgid",
1400                 .data           = &fs_overflowgid,
1401                 .maxlen         = sizeof(int),
1402                 .mode           = 0644,
1403                 .proc_handler   = proc_dointvec_minmax,
1404                 .extra1         = &minolduid,
1405                 .extra2         = &maxolduid,
1406         },
1407 #ifdef CONFIG_FILE_LOCKING
1408         {
1409                 .procname       = "leases-enable",
1410                 .data           = &leases_enable,
1411                 .maxlen         = sizeof(int),
1412                 .mode           = 0644,
1413                 .proc_handler   = proc_dointvec,
1414         },
1415 #endif
1416 #ifdef CONFIG_DNOTIFY
1417         {
1418                 .procname       = "dir-notify-enable",
1419                 .data           = &dir_notify_enable,
1420                 .maxlen         = sizeof(int),
1421                 .mode           = 0644,
1422                 .proc_handler   = proc_dointvec,
1423         },
1424 #endif
1425 #ifdef CONFIG_MMU
1426 #ifdef CONFIG_FILE_LOCKING
1427         {
1428                 .procname       = "lease-break-time",
1429                 .data           = &lease_break_time,
1430                 .maxlen         = sizeof(int),
1431                 .mode           = 0644,
1432                 .proc_handler   = proc_dointvec,
1433         },
1434 #endif
1435 #ifdef CONFIG_AIO
1436         {
1437                 .procname       = "aio-nr",
1438                 .data           = &aio_nr,
1439                 .maxlen         = sizeof(aio_nr),
1440                 .mode           = 0444,
1441                 .proc_handler   = proc_doulongvec_minmax,
1442         },
1443         {
1444                 .procname       = "aio-max-nr",
1445                 .data           = &aio_max_nr,
1446                 .maxlen         = sizeof(aio_max_nr),
1447                 .mode           = 0644,
1448                 .proc_handler   = proc_doulongvec_minmax,
1449         },
1450 #endif /* CONFIG_AIO */
1451 #ifdef CONFIG_INOTIFY_USER
1452         {
1453                 .procname       = "inotify",
1454                 .mode           = 0555,
1455                 .child          = inotify_table,
1456         },
1457 #endif
1458 #ifdef CONFIG_EPOLL
1459         {
1460                 .procname       = "epoll",
1461                 .mode           = 0555,
1462                 .child          = epoll_table,
1463         },
1464 #endif
1465 #endif
1466         {
1467                 .procname       = "suid_dumpable",
1468                 .data           = &suid_dumpable,
1469                 .maxlen         = sizeof(int),
1470                 .mode           = 0644,
1471                 .proc_handler   = proc_dointvec_minmax,
1472                 .extra1         = &zero,
1473                 .extra2         = &two,
1474         },
1475 #if defined(CONFIG_BINFMT_MISC) || defined(CONFIG_BINFMT_MISC_MODULE)
1476         {
1477                 .procname       = "binfmt_misc",
1478                 .mode           = 0555,
1479                 .child          = binfmt_misc_table,
1480         },
1481 #endif
1482         {
1483                 .procname       = "pipe-max-size",
1484                 .data           = &pipe_max_size,
1485                 .maxlen         = sizeof(int),
1486                 .mode           = 0644,
1487                 .proc_handler   = &pipe_proc_fn,
1488                 .extra1         = &pipe_min_size,
1489         },
1490 /*
1491  * NOTE: do not add new entries to this table unless you have read
1492  * Documentation/sysctl/ctl_unnumbered.txt
1493  */
1494         { }
1495 };
1496
1497 static struct ctl_table debug_table[] = {
1498 #if defined(CONFIG_X86) || defined(CONFIG_PPC) || defined(CONFIG_SPARC) || \
1499     defined(CONFIG_S390)
1500         {
1501                 .procname       = "exception-trace",
1502                 .data           = &show_unhandled_signals,
1503                 .maxlen         = sizeof(int),
1504                 .mode           = 0644,
1505                 .proc_handler   = proc_dointvec
1506         },
1507 #endif
1508 #if defined(CONFIG_OPTPROBES)
1509         {
1510                 .procname       = "kprobes-optimization",
1511                 .data           = &sysctl_kprobes_optimization,
1512                 .maxlen         = sizeof(int),
1513                 .mode           = 0644,
1514                 .proc_handler   = proc_kprobes_optimization_handler,
1515                 .extra1         = &zero,
1516                 .extra2         = &one,
1517         },
1518 #endif
1519         { }
1520 };
1521
1522 static struct ctl_table dev_table[] = {
1523         { }
1524 };
1525
1526 static DEFINE_SPINLOCK(sysctl_lock);
1527
1528 /* called under sysctl_lock */
1529 static int use_table(struct ctl_table_header *p)
1530 {
1531         if (unlikely(p->unregistering))
1532                 return 0;
1533         p->used++;
1534         return 1;
1535 }
1536
1537 /* called under sysctl_lock */
1538 static void unuse_table(struct ctl_table_header *p)
1539 {
1540         if (!--p->used)
1541                 if (unlikely(p->unregistering))
1542                         complete(p->unregistering);
1543 }
1544
1545 /* called under sysctl_lock, will reacquire if has to wait */
1546 static void start_unregistering(struct ctl_table_header *p)
1547 {
1548         /*
1549          * if p->used is 0, nobody will ever touch that entry again;
1550          * we'll eliminate all paths to it before dropping sysctl_lock
1551          */
1552         if (unlikely(p->used)) {
1553                 struct completion wait;
1554                 init_completion(&wait);
1555                 p->unregistering = &wait;
1556                 spin_unlock(&sysctl_lock);
1557                 wait_for_completion(&wait);
1558                 spin_lock(&sysctl_lock);
1559         } else {
1560                 /* anything non-NULL; we'll never dereference it */
1561                 p->unregistering = ERR_PTR(-EINVAL);
1562         }
1563         /*
1564          * do not remove from the list until nobody holds it; walking the
1565          * list in do_sysctl() relies on that.
1566          */
1567         list_del_init(&p->ctl_entry);
1568 }
1569
1570 void sysctl_head_get(struct ctl_table_header *head)
1571 {
1572         spin_lock(&sysctl_lock);
1573         head->count++;
1574         spin_unlock(&sysctl_lock);
1575 }
1576
1577 void sysctl_head_put(struct ctl_table_header *head)
1578 {
1579         spin_lock(&sysctl_lock);
1580         if (!--head->count)
1581                 kfree(head);
1582         spin_unlock(&sysctl_lock);
1583 }
1584
1585 struct ctl_table_header *sysctl_head_grab(struct ctl_table_header *head)
1586 {
1587         if (!head)
1588                 BUG();
1589         spin_lock(&sysctl_lock);
1590         if (!use_table(head))
1591                 head = ERR_PTR(-ENOENT);
1592         spin_unlock(&sysctl_lock);
1593         return head;
1594 }
1595
1596 void sysctl_head_finish(struct ctl_table_header *head)
1597 {
1598         if (!head)
1599                 return;
1600         spin_lock(&sysctl_lock);
1601         unuse_table(head);
1602         spin_unlock(&sysctl_lock);
1603 }
1604
1605 static struct ctl_table_set *
1606 lookup_header_set(struct ctl_table_root *root, struct nsproxy *namespaces)
1607 {
1608         struct ctl_table_set *set = &root->default_set;
1609         if (root->lookup)
1610                 set = root->lookup(root, namespaces);
1611         return set;
1612 }
1613
1614 static struct list_head *
1615 lookup_header_list(struct ctl_table_root *root, struct nsproxy *namespaces)
1616 {
1617         struct ctl_table_set *set = lookup_header_set(root, namespaces);
1618         return &set->list;
1619 }
1620
1621 struct ctl_table_header *__sysctl_head_next(struct nsproxy *namespaces,
1622                                             struct ctl_table_header *prev)
1623 {
1624         struct ctl_table_root *root;
1625         struct list_head *header_list;
1626         struct ctl_table_header *head;
1627         struct list_head *tmp;
1628
1629         spin_lock(&sysctl_lock);
1630         if (prev) {
1631                 head = prev;
1632                 tmp = &prev->ctl_entry;
1633                 unuse_table(prev);
1634                 goto next;
1635         }
1636         tmp = &root_table_header.ctl_entry;
1637         for (;;) {
1638                 head = list_entry(tmp, struct ctl_table_header, ctl_entry);
1639
1640                 if (!use_table(head))
1641                         goto next;
1642                 spin_unlock(&sysctl_lock);
1643                 return head;
1644         next:
1645                 root = head->root;
1646                 tmp = tmp->next;
1647                 header_list = lookup_header_list(root, namespaces);
1648                 if (tmp != header_list)
1649                         continue;
1650
1651                 do {
1652                         root = list_entry(root->root_list.next,
1653                                         struct ctl_table_root, root_list);
1654                         if (root == &sysctl_table_root)
1655                                 goto out;
1656                         header_list = lookup_header_list(root, namespaces);
1657                 } while (list_empty(header_list));
1658                 tmp = header_list->next;
1659         }
1660 out:
1661         spin_unlock(&sysctl_lock);
1662         return NULL;
1663 }
1664
1665 struct ctl_table_header *sysctl_head_next(struct ctl_table_header *prev)
1666 {
1667         return __sysctl_head_next(current->nsproxy, prev);
1668 }
1669
1670 void register_sysctl_root(struct ctl_table_root *root)
1671 {
1672         spin_lock(&sysctl_lock);
1673         list_add_tail(&root->root_list, &sysctl_table_root.root_list);
1674         spin_unlock(&sysctl_lock);
1675 }
1676
1677 /*
1678  * sysctl_perm does NOT grant the superuser all rights automatically, because
1679  * some sysctl variables are readonly even to root.
1680  */
1681
1682 static int test_perm(int mode, int op)
1683 {
1684         if (!current_euid())
1685                 mode >>= 6;
1686         else if (in_egroup_p(0))
1687                 mode >>= 3;
1688         if ((op & ~mode & (MAY_READ|MAY_WRITE|MAY_EXEC)) == 0)
1689                 return 0;
1690         return -EACCES;
1691 }
1692
1693 int sysctl_perm(struct ctl_table_root *root, struct ctl_table *table, int op)
1694 {
1695         int error;
1696         int mode;
1697
1698         error = security_sysctl(table, op & (MAY_READ | MAY_WRITE | MAY_EXEC));
1699         if (error)
1700                 return error;
1701
1702         if (root->permissions)
1703                 mode = root->permissions(root, current->nsproxy, table);
1704         else
1705                 mode = table->mode;
1706
1707         return test_perm(mode, op);
1708 }
1709
1710 static void sysctl_set_parent(struct ctl_table *parent, struct ctl_table *table)
1711 {
1712         for (; table->procname; table++) {
1713                 table->parent = parent;
1714                 if (table->child)
1715                         sysctl_set_parent(table, table->child);
1716         }
1717 }
1718
1719 static __init int sysctl_init(void)
1720 {
1721         sysctl_set_parent(NULL, root_table);
1722 #ifdef CONFIG_SYSCTL_SYSCALL_CHECK
1723         sysctl_check_table(current->nsproxy, root_table);
1724 #endif
1725         return 0;
1726 }
1727
1728 core_initcall(sysctl_init);
1729
1730 static struct ctl_table *is_branch_in(struct ctl_table *branch,
1731                                       struct ctl_table *table)
1732 {
1733         struct ctl_table *p;
1734         const char *s = branch->procname;
1735
1736         /* branch should have named subdirectory as its first element */
1737         if (!s || !branch->child)
1738                 return NULL;
1739
1740         /* ... and nothing else */
1741         if (branch[1].procname)
1742                 return NULL;
1743
1744         /* table should contain subdirectory with the same name */
1745         for (p = table; p->procname; p++) {
1746                 if (!p->child)
1747                         continue;
1748                 if (p->procname && strcmp(p->procname, s) == 0)
1749                         return p;
1750         }
1751         return NULL;
1752 }
1753
1754 /* see if attaching q to p would be an improvement */
1755 static void try_attach(struct ctl_table_header *p, struct ctl_table_header *q)
1756 {
1757         struct ctl_table *to = p->ctl_table, *by = q->ctl_table;
1758         struct ctl_table *next;
1759         int is_better = 0;
1760         int not_in_parent = !p->attached_by;
1761
1762         while ((next = is_branch_in(by, to)) != NULL) {
1763                 if (by == q->attached_by)
1764                         is_better = 1;
1765                 if (to == p->attached_by)
1766                         not_in_parent = 1;
1767                 by = by->child;
1768                 to = next->child;
1769         }
1770
1771         if (is_better && not_in_parent) {
1772                 q->attached_by = by;
1773                 q->attached_to = to;
1774                 q->parent = p;
1775         }
1776 }
1777
1778 /**
1779  * __register_sysctl_paths - register a sysctl hierarchy
1780  * @root: List of sysctl headers to register on
1781  * @namespaces: Data to compute which lists of sysctl entries are visible
1782  * @path: The path to the directory the sysctl table is in.
1783  * @table: the top-level table structure
1784  *
1785  * Register a sysctl table hierarchy. @table should be a filled in ctl_table
1786  * array. A completely 0 filled entry terminates the table.
1787  *
1788  * The members of the &struct ctl_table structure are used as follows:
1789  *
1790  * procname - the name of the sysctl file under /proc/sys. Set to %NULL to not
1791  *            enter a sysctl file
1792  *
1793  * data - a pointer to data for use by proc_handler
1794  *
1795  * maxlen - the maximum size in bytes of the data
1796  *
1797  * mode - the file permissions for the /proc/sys file, and for sysctl(2)
1798  *
1799  * child - a pointer to the child sysctl table if this entry is a directory, or
1800  *         %NULL.
1801  *
1802  * proc_handler - the text handler routine (described below)
1803  *
1804  * de - for internal use by the sysctl routines
1805  *
1806  * extra1, extra2 - extra pointers usable by the proc handler routines
1807  *
1808  * Leaf nodes in the sysctl tree will be represented by a single file
1809  * under /proc; non-leaf nodes will be represented by directories.
1810  *
1811  * sysctl(2) can automatically manage read and write requests through
1812  * the sysctl table.  The data and maxlen fields of the ctl_table
1813  * struct enable minimal validation of the values being written to be
1814  * performed, and the mode field allows minimal authentication.
1815  *
1816  * There must be a proc_handler routine for any terminal nodes
1817  * mirrored under /proc/sys (non-terminals are handled by a built-in
1818  * directory handler).  Several default handlers are available to
1819  * cover common cases -
1820  *
1821  * proc_dostring(), proc_dointvec(), proc_dointvec_jiffies(),
1822  * proc_dointvec_userhz_jiffies(), proc_dointvec_minmax(),
1823  * proc_doulongvec_ms_jiffies_minmax(), proc_doulongvec_minmax()
1824  *
1825  * It is the handler's job to read the input buffer from user memory
1826  * and process it. The handler should return 0 on success.
1827  *
1828  * This routine returns %NULL on a failure to register, and a pointer
1829  * to the table header on success.
1830  */
1831 struct ctl_table_header *__register_sysctl_paths(
1832         struct ctl_table_root *root,
1833         struct nsproxy *namespaces,
1834         const struct ctl_path *path, struct ctl_table *table)
1835 {
1836         struct ctl_table_header *header;
1837         struct ctl_table *new, **prevp;
1838         unsigned int n, npath;
1839         struct ctl_table_set *set;
1840
1841         /* Count the path components */
1842         for (npath = 0; path[npath].procname; ++npath)
1843                 ;
1844
1845         /*
1846          * For each path component, allocate a 2-element ctl_table array.
1847          * The first array element will be filled with the sysctl entry
1848          * for this, the second will be the sentinel (procname == 0).
1849          *
1850          * We allocate everything in one go so that we don't have to
1851          * worry about freeing additional memory in unregister_sysctl_table.
1852          */
1853         header = kzalloc(sizeof(struct ctl_table_header) +
1854                          (2 * npath * sizeof(struct ctl_table)), GFP_KERNEL);
1855         if (!header)
1856                 return NULL;
1857
1858         new = (struct ctl_table *) (header + 1);
1859
1860         /* Now connect the dots */
1861         prevp = &header->ctl_table;
1862         for (n = 0; n < npath; ++n, ++path) {
1863                 /* Copy the procname */
1864                 new->procname = path->procname;
1865                 new->mode     = 0555;
1866
1867                 *prevp = new;
1868                 prevp = &new->child;
1869
1870                 new += 2;
1871         }
1872         *prevp = table;
1873         header->ctl_table_arg = table;
1874
1875         INIT_LIST_HEAD(&header->ctl_entry);
1876         header->used = 0;
1877         header->unregistering = NULL;
1878         header->root = root;
1879         sysctl_set_parent(NULL, header->ctl_table);
1880         header->count = 1;
1881 #ifdef CONFIG_SYSCTL_SYSCALL_CHECK
1882         if (sysctl_check_table(namespaces, header->ctl_table)) {
1883                 kfree(header);
1884                 return NULL;
1885         }
1886 #endif
1887         spin_lock(&sysctl_lock);
1888         header->set = lookup_header_set(root, namespaces);
1889         header->attached_by = header->ctl_table;
1890         header->attached_to = root_table;
1891         header->parent = &root_table_header;
1892         for (set = header->set; set; set = set->parent) {
1893                 struct ctl_table_header *p;
1894                 list_for_each_entry(p, &set->list, ctl_entry) {
1895                         if (p->unregistering)
1896                                 continue;
1897                         try_attach(p, header);
1898                 }
1899         }
1900         header->parent->count++;
1901         list_add_tail(&header->ctl_entry, &header->set->list);
1902         spin_unlock(&sysctl_lock);
1903
1904         return header;
1905 }
1906
1907 /**
1908  * register_sysctl_table_path - register a sysctl table hierarchy
1909  * @path: The path to the directory the sysctl table is in.
1910  * @table: the top-level table structure
1911  *
1912  * Register a sysctl table hierarchy. @table should be a filled in ctl_table
1913  * array. A completely 0 filled entry terminates the table.
1914  *
1915  * See __register_sysctl_paths for more details.
1916  */
1917 struct ctl_table_header *register_sysctl_paths(const struct ctl_path *path,
1918                                                 struct ctl_table *table)
1919 {
1920         return __register_sysctl_paths(&sysctl_table_root, current->nsproxy,
1921                                         path, table);
1922 }
1923
1924 /**
1925  * register_sysctl_table - register a sysctl table hierarchy
1926  * @table: the top-level table structure
1927  *
1928  * Register a sysctl table hierarchy. @table should be a filled in ctl_table
1929  * array. A completely 0 filled entry terminates the table.
1930  *
1931  * See register_sysctl_paths for more details.
1932  */
1933 struct ctl_table_header *register_sysctl_table(struct ctl_table *table)
1934 {
1935         static const struct ctl_path null_path[] = { {} };
1936
1937         return register_sysctl_paths(null_path, table);
1938 }
1939
1940 /**
1941  * unregister_sysctl_table - unregister a sysctl table hierarchy
1942  * @header: the header returned from register_sysctl_table
1943  *
1944  * Unregisters the sysctl table and all children. proc entries may not
1945  * actually be removed until they are no longer used by anyone.
1946  */
1947 void unregister_sysctl_table(struct ctl_table_header * header)
1948 {
1949         might_sleep();
1950
1951         if (header == NULL)
1952                 return;
1953
1954         spin_lock(&sysctl_lock);
1955         start_unregistering(header);
1956         if (!--header->parent->count) {
1957                 WARN_ON(1);
1958                 kfree(header->parent);
1959         }
1960         if (!--header->count)
1961                 kfree(header);
1962         spin_unlock(&sysctl_lock);
1963 }
1964
1965 int sysctl_is_seen(struct ctl_table_header *p)
1966 {
1967         struct ctl_table_set *set = p->set;
1968         int res;
1969         spin_lock(&sysctl_lock);
1970         if (p->unregistering)
1971                 res = 0;
1972         else if (!set->is_seen)
1973                 res = 1;
1974         else
1975                 res = set->is_seen(set);
1976         spin_unlock(&sysctl_lock);
1977         return res;
1978 }
1979
1980 void setup_sysctl_set(struct ctl_table_set *p,
1981         struct ctl_table_set *parent,
1982         int (*is_seen)(struct ctl_table_set *))
1983 {
1984         INIT_LIST_HEAD(&p->list);
1985         p->parent = parent ? parent : &sysctl_table_root.default_set;
1986         p->is_seen = is_seen;
1987 }
1988
1989 #else /* !CONFIG_SYSCTL */
1990 struct ctl_table_header *register_sysctl_table(struct ctl_table * table)
1991 {
1992         return NULL;
1993 }
1994
1995 struct ctl_table_header *register_sysctl_paths(const struct ctl_path *path,
1996                                                     struct ctl_table *table)
1997 {
1998         return NULL;
1999 }
2000
2001 void unregister_sysctl_table(struct ctl_table_header * table)
2002 {
2003 }
2004
2005 void setup_sysctl_set(struct ctl_table_set *p,
2006         struct ctl_table_set *parent,
2007         int (*is_seen)(struct ctl_table_set *))
2008 {
2009 }
2010
2011 void sysctl_head_put(struct ctl_table_header *head)
2012 {
2013 }
2014
2015 #endif /* CONFIG_SYSCTL */
2016
2017 /*
2018  * /proc/sys support
2019  */
2020
2021 #ifdef CONFIG_PROC_SYSCTL
2022
2023 static int _proc_do_string(void* data, int maxlen, int write,
2024                            void __user *buffer,
2025                            size_t *lenp, loff_t *ppos)
2026 {
2027         size_t len;
2028         char __user *p;
2029         char c;
2030
2031         if (!data || !maxlen || !*lenp) {
2032                 *lenp = 0;
2033                 return 0;
2034         }
2035
2036         if (write) {
2037                 len = 0;
2038                 p = buffer;
2039                 while (len < *lenp) {
2040                         if (get_user(c, p++))
2041                                 return -EFAULT;
2042                         if (c == 0 || c == '\n')
2043                                 break;
2044                         len++;
2045                 }
2046                 if (len >= maxlen)
2047                         len = maxlen-1;
2048                 if(copy_from_user(data, buffer, len))
2049                         return -EFAULT;
2050                 ((char *) data)[len] = 0;
2051                 *ppos += *lenp;
2052         } else {
2053                 len = strlen(data);
2054                 if (len > maxlen)
2055                         len = maxlen;
2056
2057                 if (*ppos > len) {
2058                         *lenp = 0;
2059                         return 0;
2060                 }
2061
2062                 data += *ppos;
2063                 len  -= *ppos;
2064
2065                 if (len > *lenp)
2066                         len = *lenp;
2067                 if (len)
2068                         if(copy_to_user(buffer, data, len))
2069                                 return -EFAULT;
2070                 if (len < *lenp) {
2071                         if(put_user('\n', ((char __user *) buffer) + len))
2072                                 return -EFAULT;
2073                         len++;
2074                 }
2075                 *lenp = len;
2076                 *ppos += len;
2077         }
2078         return 0;
2079 }
2080
2081 /**
2082  * proc_dostring - read a string sysctl
2083  * @table: the sysctl table
2084  * @write: %TRUE if this is a write to the sysctl file
2085  * @buffer: the user buffer
2086  * @lenp: the size of the user buffer
2087  * @ppos: file position
2088  *
2089  * Reads/writes a string from/to the user buffer. If the kernel
2090  * buffer provided is not large enough to hold the string, the
2091  * string is truncated. The copied string is %NULL-terminated.
2092  * If the string is being read by the user process, it is copied
2093  * and a newline '\n' is added. It is truncated if the buffer is
2094  * not large enough.
2095  *
2096  * Returns 0 on success.
2097  */
2098 int proc_dostring(struct ctl_table *table, int write,
2099                   void __user *buffer, size_t *lenp, loff_t *ppos)
2100 {
2101         return _proc_do_string(table->data, table->maxlen, write,
2102                                buffer, lenp, ppos);
2103 }
2104
2105 static size_t proc_skip_spaces(char **buf)
2106 {
2107         size_t ret;
2108         char *tmp = skip_spaces(*buf);
2109         ret = tmp - *buf;
2110         *buf = tmp;
2111         return ret;
2112 }
2113
2114 static void proc_skip_char(char **buf, size_t *size, const char v)
2115 {
2116         while (*size) {
2117                 if (**buf != v)
2118                         break;
2119                 (*size)--;
2120                 (*buf)++;
2121         }
2122 }
2123
2124 #define TMPBUFLEN 22
2125 /**
2126  * proc_get_long - reads an ASCII formatted integer from a user buffer
2127  *
2128  * @buf: a kernel buffer
2129  * @size: size of the kernel buffer
2130  * @val: this is where the number will be stored
2131  * @neg: set to %TRUE if number is negative
2132  * @perm_tr: a vector which contains the allowed trailers
2133  * @perm_tr_len: size of the perm_tr vector
2134  * @tr: pointer to store the trailer character
2135  *
2136  * In case of success %0 is returned and @buf and @size are updated with
2137  * the amount of bytes read. If @tr is non-NULL and a trailing
2138  * character exists (size is non-zero after returning from this
2139  * function), @tr is updated with the trailing character.
2140  */
2141 static int proc_get_long(char **buf, size_t *size,
2142                           unsigned long *val, bool *neg,
2143                           const char *perm_tr, unsigned perm_tr_len, char *tr)
2144 {
2145         int len;
2146         char *p, tmp[TMPBUFLEN];
2147
2148         if (!*size)
2149                 return -EINVAL;
2150
2151         len = *size;
2152         if (len > TMPBUFLEN - 1)
2153                 len = TMPBUFLEN - 1;
2154
2155         memcpy(tmp, *buf, len);
2156
2157         tmp[len] = 0;
2158         p = tmp;
2159         if (*p == '-' && *size > 1) {
2160                 *neg = true;
2161                 p++;
2162         } else
2163                 *neg = false;
2164         if (!isdigit(*p))
2165                 return -EINVAL;
2166
2167         *val = simple_strtoul(p, &p, 0);
2168
2169         len = p - tmp;
2170
2171         /* We don't know if the next char is whitespace thus we may accept
2172          * invalid integers (e.g. 1234...a) or two integers instead of one
2173          * (e.g. 123...1). So lets not allow such large numbers. */
2174         if (len == TMPBUFLEN - 1)
2175                 return -EINVAL;
2176
2177         if (len < *size && perm_tr_len && !memchr(perm_tr, *p, perm_tr_len))
2178                 return -EINVAL;
2179
2180         if (tr && (len < *size))
2181                 *tr = *p;
2182
2183         *buf += len;
2184         *size -= len;
2185
2186         return 0;
2187 }
2188
2189 /**
2190  * proc_put_long - converts an integer to a decimal ASCII formatted string
2191  *
2192  * @buf: the user buffer
2193  * @size: the size of the user buffer
2194  * @val: the integer to be converted
2195  * @neg: sign of the number, %TRUE for negative
2196  *
2197  * In case of success %0 is returned and @buf and @size are updated with
2198  * the amount of bytes written.
2199  */
2200 static int proc_put_long(void __user **buf, size_t *size, unsigned long val,
2201                           bool neg)
2202 {
2203         int len;
2204         char tmp[TMPBUFLEN], *p = tmp;
2205
2206         sprintf(p, "%s%lu", neg ? "-" : "", val);
2207         len = strlen(tmp);
2208         if (len > *size)
2209                 len = *size;
2210         if (copy_to_user(*buf, tmp, len))
2211                 return -EFAULT;
2212         *size -= len;
2213         *buf += len;
2214         return 0;
2215 }
2216 #undef TMPBUFLEN
2217
2218 static int proc_put_char(void __user **buf, size_t *size, char c)
2219 {
2220         if (*size) {
2221                 char __user **buffer = (char __user **)buf;
2222                 if (put_user(c, *buffer))
2223                         return -EFAULT;
2224                 (*size)--, (*buffer)++;
2225                 *buf = *buffer;
2226         }
2227         return 0;
2228 }
2229
2230 static int do_proc_dointvec_conv(bool *negp, unsigned long *lvalp,
2231                                  int *valp,
2232                                  int write, void *data)
2233 {
2234         if (write) {
2235                 *valp = *negp ? -*lvalp : *lvalp;
2236         } else {
2237                 int val = *valp;
2238                 if (val < 0) {
2239                         *negp = true;
2240                         *lvalp = (unsigned long)-val;
2241                 } else {
2242                         *negp = false;
2243                         *lvalp = (unsigned long)val;
2244                 }
2245         }
2246         return 0;
2247 }
2248
2249 static const char proc_wspace_sep[] = { ' ', '\t', '\n' };
2250
2251 static int __do_proc_dointvec(void *tbl_data, struct ctl_table *table,
2252                   int write, void __user *buffer,
2253                   size_t *lenp, loff_t *ppos,
2254                   int (*conv)(bool *negp, unsigned long *lvalp, int *valp,
2255                               int write, void *data),
2256                   void *data)
2257 {
2258         int *i, vleft, first = 1, err = 0;
2259         unsigned long page = 0;
2260         size_t left;
2261         char *kbuf;
2262
2263         if (!tbl_data || !table->maxlen || !*lenp || (*ppos && !write)) {
2264                 *lenp = 0;
2265                 return 0;
2266         }
2267
2268         i = (int *) tbl_data;
2269         vleft = table->maxlen / sizeof(*i);
2270         left = *lenp;
2271
2272         if (!conv)
2273                 conv = do_proc_dointvec_conv;
2274
2275         if (write) {
2276                 if (left > PAGE_SIZE - 1)
2277                         left = PAGE_SIZE - 1;
2278                 page = __get_free_page(GFP_TEMPORARY);
2279                 kbuf = (char *) page;
2280                 if (!kbuf)
2281                         return -ENOMEM;
2282                 if (copy_from_user(kbuf, buffer, left)) {
2283                         err = -EFAULT;
2284                         goto free;
2285                 }
2286                 kbuf[left] = 0;
2287         }
2288
2289         for (; left && vleft--; i++, first=0) {
2290                 unsigned long lval;
2291                 bool neg;
2292
2293                 if (write) {
2294                         left -= proc_skip_spaces(&kbuf);
2295
2296                         if (!left)
2297                                 break;
2298                         err = proc_get_long(&kbuf, &left, &lval, &neg,
2299                                              proc_wspace_sep,
2300                                              sizeof(proc_wspace_sep), NULL);
2301                         if (err)
2302                                 break;
2303                         if (conv(&neg, &lval, i, 1, data)) {
2304                                 err = -EINVAL;
2305                                 break;
2306                         }
2307                 } else {
2308                         if (conv(&neg, &lval, i, 0, data)) {
2309                                 err = -EINVAL;
2310                                 break;
2311                         }
2312                         if (!first)
2313                                 err = proc_put_char(&buffer, &left, '\t');
2314                         if (err)
2315                                 break;
2316                         err = proc_put_long(&buffer, &left, lval, neg);
2317                         if (err)
2318                                 break;
2319                 }
2320         }
2321
2322         if (!write && !first && left && !err)
2323                 err = proc_put_char(&buffer, &left, '\n');
2324         if (write && !err && left)
2325                 left -= proc_skip_spaces(&kbuf);
2326 free:
2327         if (write) {
2328                 free_page(page);
2329                 if (first)
2330                         return err ? : -EINVAL;
2331         }
2332         *lenp -= left;
2333         *ppos += *lenp;
2334         return err;
2335 }
2336
2337 static int do_proc_dointvec(struct ctl_table *table, int write,
2338                   void __user *buffer, size_t *lenp, loff_t *ppos,
2339                   int (*conv)(bool *negp, unsigned long *lvalp, int *valp,
2340                               int write, void *data),
2341                   void *data)
2342 {
2343         return __do_proc_dointvec(table->data, table, write,
2344                         buffer, lenp, ppos, conv, data);
2345 }
2346
2347 /**
2348  * proc_dointvec - read a vector of integers
2349  * @table: the sysctl table
2350  * @write: %TRUE if this is a write to the sysctl file
2351  * @buffer: the user buffer
2352  * @lenp: the size of the user buffer
2353  * @ppos: file position
2354  *
2355  * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2356  * values from/to the user buffer, treated as an ASCII string.
2357  *
2358  * Returns 0 on success.
2359  */
2360 int proc_dointvec(struct ctl_table *table, int write,
2361                      void __user *buffer, size_t *lenp, loff_t *ppos)
2362 {
2363     return do_proc_dointvec(table,write,buffer,lenp,ppos,
2364                             NULL,NULL);
2365 }
2366
2367 /*
2368  * Taint values can only be increased
2369  * This means we can safely use a temporary.
2370  */
2371 static int proc_taint(struct ctl_table *table, int write,
2372                                void __user *buffer, size_t *lenp, loff_t *ppos)
2373 {
2374         struct ctl_table t;
2375         unsigned long tmptaint = get_taint();
2376         int err;
2377
2378         if (write && !capable(CAP_SYS_ADMIN))
2379                 return -EPERM;
2380
2381         t = *table;
2382         t.data = &tmptaint;
2383         err = proc_doulongvec_minmax(&t, write, buffer, lenp, ppos);
2384         if (err < 0)
2385                 return err;
2386
2387         if (write) {
2388                 /*
2389                  * Poor man's atomic or. Not worth adding a primitive
2390                  * to everyone's atomic.h for this
2391                  */
2392                 int i;
2393                 for (i = 0; i < BITS_PER_LONG && tmptaint >> i; i++) {
2394                         if ((tmptaint >> i) & 1)
2395                                 add_taint(i);
2396                 }
2397         }
2398
2399         return err;
2400 }
2401
2402 struct do_proc_dointvec_minmax_conv_param {
2403         int *min;
2404         int *max;
2405 };
2406
2407 static int do_proc_dointvec_minmax_conv(bool *negp, unsigned long *lvalp,
2408                                         int *valp,
2409                                         int write, void *data)
2410 {
2411         struct do_proc_dointvec_minmax_conv_param *param = data;
2412         if (write) {
2413                 int val = *negp ? -*lvalp : *lvalp;
2414                 if ((param->min && *param->min > val) ||
2415                     (param->max && *param->max < val))
2416                         return -EINVAL;
2417                 *valp = val;
2418         } else {
2419                 int val = *valp;
2420                 if (val < 0) {
2421                         *negp = true;
2422                         *lvalp = (unsigned long)-val;
2423                 } else {
2424                         *negp = false;
2425                         *lvalp = (unsigned long)val;
2426                 }
2427         }
2428         return 0;
2429 }
2430
2431 /**
2432  * proc_dointvec_minmax - read a vector of integers with min/max values
2433  * @table: the sysctl table
2434  * @write: %TRUE if this is a write to the sysctl file
2435  * @buffer: the user buffer
2436  * @lenp: the size of the user buffer
2437  * @ppos: file position
2438  *
2439  * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2440  * values from/to the user buffer, treated as an ASCII string.
2441  *
2442  * This routine will ensure the values are within the range specified by
2443  * table->extra1 (min) and table->extra2 (max).
2444  *
2445  * Returns 0 on success.
2446  */
2447 int proc_dointvec_minmax(struct ctl_table *table, int write,
2448                   void __user *buffer, size_t *lenp, loff_t *ppos)
2449 {
2450         struct do_proc_dointvec_minmax_conv_param param = {
2451                 .min = (int *) table->extra1,
2452                 .max = (int *) table->extra2,
2453         };
2454         return do_proc_dointvec(table, write, buffer, lenp, ppos,
2455                                 do_proc_dointvec_minmax_conv, &param);
2456 }
2457
2458 static int __do_proc_doulongvec_minmax(void *data, struct ctl_table *table, int write,
2459                                      void __user *buffer,
2460                                      size_t *lenp, loff_t *ppos,
2461                                      unsigned long convmul,
2462                                      unsigned long convdiv)
2463 {
2464         unsigned long *i, *min, *max;
2465         int vleft, first = 1, err = 0;
2466         unsigned long page = 0;
2467         size_t left;
2468         char *kbuf;
2469
2470         if (!data || !table->maxlen || !*lenp || (*ppos && !write)) {
2471                 *lenp = 0;
2472                 return 0;
2473         }
2474
2475         i = (unsigned long *) data;
2476         min = (unsigned long *) table->extra1;
2477         max = (unsigned long *) table->extra2;
2478         vleft = table->maxlen / sizeof(unsigned long);
2479         left = *lenp;
2480
2481         if (write) {
2482                 if (left > PAGE_SIZE - 1)
2483                         left = PAGE_SIZE - 1;
2484                 page = __get_free_page(GFP_TEMPORARY);
2485                 kbuf = (char *) page;
2486                 if (!kbuf)
2487                         return -ENOMEM;
2488                 if (copy_from_user(kbuf, buffer, left)) {
2489                         err = -EFAULT;
2490                         goto free;
2491                 }
2492                 kbuf[left] = 0;
2493         }
2494
2495         for (; left && vleft--; i++, first = 0) {
2496                 unsigned long val;
2497
2498                 if (write) {
2499                         bool neg;
2500
2501                         left -= proc_skip_spaces(&kbuf);
2502
2503                         err = proc_get_long(&kbuf, &left, &val, &neg,
2504                                              proc_wspace_sep,
2505                                              sizeof(proc_wspace_sep), NULL);
2506                         if (err)
2507                                 break;
2508                         if (neg)
2509                                 continue;
2510                         if ((min && val < *min) || (max && val > *max))
2511                                 continue;
2512                         *i = val;
2513                 } else {
2514                         val = convdiv * (*i) / convmul;
2515                         if (!first)
2516                                 err = proc_put_char(&buffer, &left, '\t');
2517                         err = proc_put_long(&buffer, &left, val, false);
2518                         if (err)
2519                                 break;
2520                 }
2521         }
2522
2523         if (!write && !first && left && !err)
2524                 err = proc_put_char(&buffer, &left, '\n');
2525         if (write && !err)
2526                 left -= proc_skip_spaces(&kbuf);
2527 free:
2528         if (write) {
2529                 free_page(page);
2530                 if (first)
2531                         return err ? : -EINVAL;
2532         }
2533         *lenp -= left;
2534         *ppos += *lenp;
2535         return err;
2536 }
2537
2538 static int do_proc_doulongvec_minmax(struct ctl_table *table, int write,
2539                                      void __user *buffer,
2540                                      size_t *lenp, loff_t *ppos,
2541                                      unsigned long convmul,
2542                                      unsigned long convdiv)
2543 {
2544         return __do_proc_doulongvec_minmax(table->data, table, write,
2545                         buffer, lenp, ppos, convmul, convdiv);
2546 }
2547
2548 /**
2549  * proc_doulongvec_minmax - read a vector of long integers with min/max values
2550  * @table: the sysctl table
2551  * @write: %TRUE if this is a write to the sysctl file
2552  * @buffer: the user buffer
2553  * @lenp: the size of the user buffer
2554  * @ppos: file position
2555  *
2556  * Reads/writes up to table->maxlen/sizeof(unsigned long) unsigned long
2557  * values from/to the user buffer, treated as an ASCII string.
2558  *
2559  * This routine will ensure the values are within the range specified by
2560  * table->extra1 (min) and table->extra2 (max).
2561  *
2562  * Returns 0 on success.
2563  */
2564 int proc_doulongvec_minmax(struct ctl_table *table, int write,
2565                            void __user *buffer, size_t *lenp, loff_t *ppos)
2566 {
2567     return do_proc_doulongvec_minmax(table, write, buffer, lenp, ppos, 1l, 1l);
2568 }
2569
2570 /**
2571  * proc_doulongvec_ms_jiffies_minmax - read a vector of millisecond values with min/max values
2572  * @table: the sysctl table
2573  * @write: %TRUE if this is a write to the sysctl file
2574  * @buffer: the user buffer
2575  * @lenp: the size of the user buffer
2576  * @ppos: file position
2577  *
2578  * Reads/writes up to table->maxlen/sizeof(unsigned long) unsigned long
2579  * values from/to the user buffer, treated as an ASCII string. The values
2580  * are treated as milliseconds, and converted to jiffies when they are stored.
2581  *
2582  * This routine will ensure the values are within the range specified by
2583  * table->extra1 (min) and table->extra2 (max).
2584  *
2585  * Returns 0 on success.
2586  */
2587 int proc_doulongvec_ms_jiffies_minmax(struct ctl_table *table, int write,
2588                                       void __user *buffer,
2589                                       size_t *lenp, loff_t *ppos)
2590 {
2591     return do_proc_doulongvec_minmax(table, write, buffer,
2592                                      lenp, ppos, HZ, 1000l);
2593 }
2594
2595
2596 static int do_proc_dointvec_jiffies_conv(bool *negp, unsigned long *lvalp,
2597                                          int *valp,
2598                                          int write, void *data)
2599 {
2600         if (write) {
2601                 if (*lvalp > LONG_MAX / HZ)
2602                         return 1;
2603                 *valp = *negp ? -(*lvalp*HZ) : (*lvalp*HZ);
2604         } else {
2605                 int val = *valp;
2606                 unsigned long lval;
2607                 if (val < 0) {
2608                         *negp = true;
2609                         lval = (unsigned long)-val;
2610                 } else {
2611                         *negp = false;
2612                         lval = (unsigned long)val;
2613                 }
2614                 *lvalp = lval / HZ;
2615         }
2616         return 0;
2617 }
2618
2619 static int do_proc_dointvec_userhz_jiffies_conv(bool *negp, unsigned long *lvalp,
2620                                                 int *valp,
2621                                                 int write, void *data)
2622 {
2623         if (write) {
2624                 if (USER_HZ < HZ && *lvalp > (LONG_MAX / HZ) * USER_HZ)
2625                         return 1;
2626                 *valp = clock_t_to_jiffies(*negp ? -*lvalp : *lvalp);
2627         } else {
2628                 int val = *valp;
2629                 unsigned long lval;
2630                 if (val < 0) {
2631                         *negp = true;
2632                         lval = (unsigned long)-val;
2633                 } else {
2634                         *negp = false;
2635                         lval = (unsigned long)val;
2636                 }
2637                 *lvalp = jiffies_to_clock_t(lval);
2638         }
2639         return 0;
2640 }
2641
2642 static int do_proc_dointvec_ms_jiffies_conv(bool *negp, unsigned long *lvalp,
2643                                             int *valp,
2644                                             int write, void *data)
2645 {
2646         if (write) {
2647                 *valp = msecs_to_jiffies(*negp ? -*lvalp : *lvalp);
2648         } else {
2649                 int val = *valp;
2650                 unsigned long lval;
2651                 if (val < 0) {
2652                         *negp = true;
2653                         lval = (unsigned long)-val;
2654                 } else {
2655                         *negp = false;
2656                         lval = (unsigned long)val;
2657                 }
2658                 *lvalp = jiffies_to_msecs(lval);
2659         }
2660         return 0;
2661 }
2662
2663 /**
2664  * proc_dointvec_jiffies - read a vector of integers as seconds
2665  * @table: the sysctl table
2666  * @write: %TRUE if this is a write to the sysctl file
2667  * @buffer: the user buffer
2668  * @lenp: the size of the user buffer
2669  * @ppos: file position
2670  *
2671  * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2672  * values from/to the user buffer, treated as an ASCII string.
2673  * The values read are assumed to be in seconds, and are converted into
2674  * jiffies.
2675  *
2676  * Returns 0 on success.
2677  */
2678 int proc_dointvec_jiffies(struct ctl_table *table, int write,
2679                           void __user *buffer, size_t *lenp, loff_t *ppos)
2680 {
2681     return do_proc_dointvec(table,write,buffer,lenp,ppos,
2682                             do_proc_dointvec_jiffies_conv,NULL);
2683 }
2684
2685 /**
2686  * proc_dointvec_userhz_jiffies - read a vector of integers as 1/USER_HZ seconds
2687  * @table: the sysctl table
2688  * @write: %TRUE if this is a write to the sysctl file
2689  * @buffer: the user buffer
2690  * @lenp: the size of the user buffer
2691  * @ppos: pointer to the file position
2692  *
2693  * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2694  * values from/to the user buffer, treated as an ASCII string.
2695  * The values read are assumed to be in 1/USER_HZ seconds, and
2696  * are converted into jiffies.
2697  *
2698  * Returns 0 on success.
2699  */
2700 int proc_dointvec_userhz_jiffies(struct ctl_table *table, int write,
2701                                  void __user *buffer, size_t *lenp, loff_t *ppos)
2702 {
2703     return do_proc_dointvec(table,write,buffer,lenp,ppos,
2704                             do_proc_dointvec_userhz_jiffies_conv,NULL);
2705 }
2706
2707 /**
2708  * proc_dointvec_ms_jiffies - read a vector of integers as 1 milliseconds
2709  * @table: the sysctl table
2710  * @write: %TRUE if this is a write to the sysctl file
2711  * @buffer: the user buffer
2712  * @lenp: the size of the user buffer
2713  * @ppos: file position
2714  * @ppos: the current position in the file
2715  *
2716  * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2717  * values from/to the user buffer, treated as an ASCII string.
2718  * The values read are assumed to be in 1/1000 seconds, and
2719  * are converted into jiffies.
2720  *
2721  * Returns 0 on success.
2722  */
2723 int proc_dointvec_ms_jiffies(struct ctl_table *table, int write,
2724                              void __user *buffer, size_t *lenp, loff_t *ppos)
2725 {
2726         return do_proc_dointvec(table, write, buffer, lenp, ppos,
2727                                 do_proc_dointvec_ms_jiffies_conv, NULL);
2728 }
2729
2730 static int proc_do_cad_pid(struct ctl_table *table, int write,
2731                            void __user *buffer, size_t *lenp, loff_t *ppos)
2732 {
2733         struct pid *new_pid;
2734         pid_t tmp;
2735         int r;
2736
2737         tmp = pid_vnr(cad_pid);
2738
2739         r = __do_proc_dointvec(&tmp, table, write, buffer,
2740                                lenp, ppos, NULL, NULL);
2741         if (r || !write)
2742                 return r;
2743
2744         new_pid = find_get_pid(tmp);
2745         if (!new_pid)
2746                 return -ESRCH;
2747
2748         put_pid(xchg(&cad_pid, new_pid));
2749         return 0;
2750 }
2751
2752 /**
2753  * proc_do_large_bitmap - read/write from/to a large bitmap
2754  * @table: the sysctl table
2755  * @write: %TRUE if this is a write to the sysctl file
2756  * @buffer: the user buffer
2757  * @lenp: the size of the user buffer
2758  * @ppos: file position
2759  *
2760  * The bitmap is stored at table->data and the bitmap length (in bits)
2761  * in table->maxlen.
2762  *
2763  * We use a range comma separated format (e.g. 1,3-4,10-10) so that
2764  * large bitmaps may be represented in a compact manner. Writing into
2765  * the file will clear the bitmap then update it with the given input.
2766  *
2767  * Returns 0 on success.
2768  */
2769 int proc_do_large_bitmap(struct ctl_table *table, int write,
2770                          void __user *buffer, size_t *lenp, loff_t *ppos)
2771 {
2772         int err = 0;
2773         bool first = 1;
2774         size_t left = *lenp;
2775         unsigned long bitmap_len = table->maxlen;
2776         unsigned long *bitmap = (unsigned long *) table->data;
2777         unsigned long *tmp_bitmap = NULL;
2778         char tr_a[] = { '-', ',', '\n' }, tr_b[] = { ',', '\n', 0 }, c;
2779
2780         if (!bitmap_len || !left || (*ppos && !write)) {
2781                 *lenp = 0;
2782                 return 0;
2783         }
2784
2785         if (write) {
2786                 unsigned long page = 0;
2787                 char *kbuf;
2788
2789                 if (left > PAGE_SIZE - 1)
2790                         left = PAGE_SIZE - 1;
2791
2792                 page = __get_free_page(GFP_TEMPORARY);
2793                 kbuf = (char *) page;
2794                 if (!kbuf)
2795                         return -ENOMEM;
2796                 if (copy_from_user(kbuf, buffer, left)) {
2797                         free_page(page);
2798                         return -EFAULT;
2799                 }
2800                 kbuf[left] = 0;
2801
2802                 tmp_bitmap = kzalloc(BITS_TO_LONGS(bitmap_len) * sizeof(unsigned long),
2803                                      GFP_KERNEL);
2804                 if (!tmp_bitmap) {
2805                         free_page(page);
2806                         return -ENOMEM;
2807                 }
2808                 proc_skip_char(&kbuf, &left, '\n');
2809                 while (!err && left) {
2810                         unsigned long val_a, val_b;
2811                         bool neg;
2812
2813                         err = proc_get_long(&kbuf, &left, &val_a, &neg, tr_a,
2814                                              sizeof(tr_a), &c);
2815                         if (err)
2816                                 break;
2817                         if (val_a >= bitmap_len || neg) {
2818                                 err = -EINVAL;
2819                                 break;
2820                         }
2821
2822                         val_b = val_a;
2823                         if (left) {
2824                                 kbuf++;
2825                                 left--;
2826                         }
2827
2828                         if (c == '-') {
2829                                 err = proc_get_long(&kbuf, &left, &val_b,
2830                                                      &neg, tr_b, sizeof(tr_b),
2831                                                      &c);
2832                                 if (err)
2833                                         break;
2834                                 if (val_b >= bitmap_len || neg ||
2835                                     val_a > val_b) {
2836                                         err = -EINVAL;
2837                                         break;
2838                                 }
2839                                 if (left) {
2840                                         kbuf++;
2841                                         left--;
2842                                 }
2843                         }
2844
2845                         while (val_a <= val_b)
2846                                 set_bit(val_a++, tmp_bitmap);
2847
2848                         first = 0;
2849                         proc_skip_char(&kbuf, &left, '\n');
2850                 }
2851                 free_page(page);
2852         } else {
2853                 unsigned long bit_a, bit_b = 0;
2854
2855                 while (left) {
2856                         bit_a = find_next_bit(bitmap, bitmap_len, bit_b);
2857                         if (bit_a >= bitmap_len)
2858                                 break;
2859                         bit_b = find_next_zero_bit(bitmap, bitmap_len,
2860                                                    bit_a + 1) - 1;
2861
2862                         if (!first) {
2863                                 err = proc_put_char(&buffer, &left, ',');
2864                                 if (err)
2865                                         break;
2866                         }
2867                         err = proc_put_long(&buffer, &left, bit_a, false);
2868                         if (err)
2869                                 break;
2870                         if (bit_a != bit_b) {
2871                                 err = proc_put_char(&buffer, &left, '-');
2872                                 if (err)
2873                                         break;
2874                                 err = proc_put_long(&buffer, &left, bit_b, false);
2875                                 if (err)
2876                                         break;
2877                         }
2878
2879                         first = 0; bit_b++;
2880                 }
2881                 if (!err)
2882                         err = proc_put_char(&buffer, &left, '\n');
2883         }
2884
2885         if (!err) {
2886                 if (write) {
2887                         if (*ppos)
2888                                 bitmap_or(bitmap, bitmap, tmp_bitmap, bitmap_len);
2889                         else
2890                                 memcpy(bitmap, tmp_bitmap,
2891                                         BITS_TO_LONGS(bitmap_len) * sizeof(unsigned long));
2892                 }
2893                 kfree(tmp_bitmap);
2894                 *lenp -= left;
2895                 *ppos += *lenp;
2896                 return 0;
2897         } else {
2898                 kfree(tmp_bitmap);
2899                 return err;
2900         }
2901 }
2902
2903 #else /* CONFIG_PROC_FS */
2904
2905 int proc_dostring(struct ctl_table *table, int write,
2906                   void __user *buffer, size_t *lenp, loff_t *ppos)
2907 {
2908         return -ENOSYS;
2909 }
2910
2911 int proc_dointvec(struct ctl_table *table, int write,
2912                   void __user *buffer, size_t *lenp, loff_t *ppos)
2913 {
2914         return -ENOSYS;
2915 }
2916
2917 int proc_dointvec_minmax(struct ctl_table *table, int write,
2918                     void __user *buffer, size_t *lenp, loff_t *ppos)
2919 {
2920         return -ENOSYS;
2921 }
2922
2923 int proc_dointvec_jiffies(struct ctl_table *table, int write,
2924                     void __user *buffer, size_t *lenp, loff_t *ppos)
2925 {
2926         return -ENOSYS;
2927 }
2928
2929 int proc_dointvec_userhz_jiffies(struct ctl_table *table, int write,
2930                     void __user *buffer, size_t *lenp, loff_t *ppos)
2931 {
2932         return -ENOSYS;
2933 }
2934
2935 int proc_dointvec_ms_jiffies(struct ctl_table *table, int write,
2936                              void __user *buffer, size_t *lenp, loff_t *ppos)
2937 {
2938         return -ENOSYS;
2939 }
2940
2941 int proc_doulongvec_minmax(struct ctl_table *table, int write,
2942                     void __user *buffer, size_t *lenp, loff_t *ppos)
2943 {
2944         return -ENOSYS;
2945 }
2946
2947 int proc_doulongvec_ms_jiffies_minmax(struct ctl_table *table, int write,
2948                                       void __user *buffer,
2949                                       size_t *lenp, loff_t *ppos)
2950 {
2951     return -ENOSYS;
2952 }
2953
2954
2955 #endif /* CONFIG_PROC_FS */
2956
2957 /*
2958  * No sense putting this after each symbol definition, twice,
2959  * exception granted :-)
2960  */
2961 EXPORT_SYMBOL(proc_dointvec);
2962 EXPORT_SYMBOL(proc_dointvec_jiffies);
2963 EXPORT_SYMBOL(proc_dointvec_minmax);
2964 EXPORT_SYMBOL(proc_dointvec_userhz_jiffies);
2965 EXPORT_SYMBOL(proc_dointvec_ms_jiffies);
2966 EXPORT_SYMBOL(proc_dostring);
2967 EXPORT_SYMBOL(proc_doulongvec_minmax);
2968 EXPORT_SYMBOL(proc_doulongvec_ms_jiffies_minmax);
2969 EXPORT_SYMBOL(register_sysctl_table);
2970 EXPORT_SYMBOL(register_sysctl_paths);
2971 EXPORT_SYMBOL(unregister_sysctl_table);