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