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 #ifdef CONFIG_LOCKUP_DETECTOR
  80 #include <linux/nmi.h>
  81 #endif
  82
  83
  84 #if defined(CONFIG_SYSCTL)
  85
  86 /* External variables not in a header file. */
  87 extern int sysctl_overcommit_memory;
  88 extern int sysctl_overcommit_ratio;
  89 extern int sysctl_panic_on_oom;
  90 extern int sysctl_oom_kill_allocating_task;
  91 extern int sysctl_oom_dump_tasks;
  92 extern int max_threads;
  93 extern int core_uses_pid;
  94 extern int suid_dumpable;
  95 extern char core_pattern[];
  96 extern unsigned int core_pipe_limit;
  97 extern int pid_max;
  98 extern int min_free_kbytes;
  99 extern int pid_max_min, pid_max_max;
 100 extern int sysctl_drop_caches;
 101 extern int percpu_pagelist_fraction;
 102 extern int compat_log;
 103 extern int latencytop_enabled;
 104 extern int sysctl_nr_open_min, sysctl_nr_open_max;
 105 #ifndef CONFIG_MMU
 106 extern int sysctl_nr_trim_pages;
 107 #endif
 108 #ifdef CONFIG_BLOCK
 109 extern int blk_iopoll_enabled;
 110 #endif
 111
 112 /* Constants used for minimum and  maximum */
 113 #ifdef CONFIG_LOCKUP_DETECTOR
 114 static int sixty = 60;
 115 static int neg_one = -1;
 116 #endif
 117
 118 static int zero;
 119 static int __maybe_unused one = 1;
 120 static int __maybe_unused two = 2;
 121 static unsigned long one_ul = 1;
 122 static int one_hundred = 100;
 123 #ifdef CONFIG_PRINTK
 124 static int ten_thousand = 10000;
 125 #endif
 126
 127 /* this is needed for the proc_doulongvec_minmax of vm_dirty_bytes */
 128 static unsigned long dirty_bytes_min = 2 * PAGE_SIZE;
 129
 130 /* this is needed for the proc_dointvec_minmax for [fs_]overflow UID and GID */
 131 static int maxolduid = 65535;
 132 static int minolduid;
 133 static int min_percpu_pagelist_fract = 8;
 134
 135 static int ngroups_max = NGROUPS_MAX;
 136
 137 #ifdef CONFIG_SPARC
 138 #include <asm/system.h>
 139 #endif
 140
 141 #ifdef CONFIG_SPARC64
 142 extern int sysctl_tsb_ratio;
 143 #endif
 144
 145 #ifdef __hppa__
 146 extern int pwrsw_enabled;
 147 extern int unaligned_enabled;
 148 #endif
 149
 150 #ifdef CONFIG_S390
 151 #ifdef CONFIG_MATHEMU
 152 extern int sysctl_ieee_emulation_warnings;
 153 #endif
 154 extern int sysctl_userprocess_debug;
 155 extern int spin_retry;
 156 #endif
 157
 158 #ifdef CONFIG_IA64
 159 extern int no_unaligned_warning;
 160 extern int unaligned_dump_stack;
 161 #endif
 162
 163 extern struct ratelimit_state printk_ratelimit_state;
 164
 165 #ifdef CONFIG_PROC_SYSCTL
 166 static int proc_do_cad_pid(struct ctl_table *table, int write,
 167                   void __user *buffer, size_t *lenp, loff_t *ppos);
 168 static int proc_taint(struct ctl_table *table, int write,
 169                                void __user *buffer, size_t *lenp, loff_t *ppos);
 170 #endif
 171
 172 #ifdef CONFIG_MAGIC_SYSRQ
 173 static int __sysrq_enabled; /* Note: sysrq code ises it's own private copy */
 174
 175 static int sysrq_sysctl_handler(ctl_table *table, int write,
 176                                 void __user *buffer, size_t *lenp,
 177                                 loff_t *ppos)
 178 {
 179         int error;
 180
 181         error = proc_dointvec(table, write, buffer, lenp, ppos);
 182         if (error)
 183                 return error;
 184
 185         if (write)
 186                 sysrq_toggle_support(__sysrq_enabled);
 187
 188         return 0;
 189 }
 190
 191 #endif
 192
 193 static struct ctl_table root_table[];
 194 static struct ctl_table_root sysctl_table_root;
 195 static struct ctl_table_header root_table_header = {
 196         .count = 1,
 197         .ctl_table = root_table,
 198         .ctl_entry = LIST_HEAD_INIT(sysctl_table_root.default_set.list),
 199         .root = &sysctl_table_root,
 200         .set = &sysctl_table_root.default_set,
 201 };
 202 static struct ctl_table_root sysctl_table_root = {
 203         .root_list = LIST_HEAD_INIT(sysctl_table_root.root_list),
 204         .default_set.list = LIST_HEAD_INIT(root_table_header.ctl_entry),
 205 };
 206
 207 static struct ctl_table kern_table[];
 208 static struct ctl_table vm_table[];
 209 static struct ctl_table fs_table[];
 210 static struct ctl_table debug_table[];
 211 static struct ctl_table dev_table[];
 212 extern struct ctl_table random_table[];
 213 #ifdef CONFIG_INOTIFY_USER
 214 extern struct ctl_table inotify_table[];
 215 #endif
 216 #ifdef CONFIG_EPOLL
 217 extern struct ctl_table epoll_table[];
 218 #endif
 219
 220 #ifdef HAVE_ARCH_PICK_MMAP_LAYOUT
 221 int sysctl_legacy_va_layout;
 222 #endif
 223
 224 /* The default sysctl tables: */
 225
 226 static struct ctl_table root_table[] = {
 227         {
 228                 .procname       = "kernel",
 229                 .mode           = 0555,
 230                 .child          = kern_table,
 231         },
 232         {
 233                 .procname       = "vm",
 234                 .mode           = 0555,
 235                 .child          = vm_table,
 236         },
 237         {
 238                 .procname       = "fs",
 239                 .mode           = 0555,
 240                 .child          = fs_table,
 241         },
 242         {
 243                 .procname       = "debug",
 244                 .mode           = 0555,
 245                 .child          = debug_table,
 246         },
 247         {
 248                 .procname       = "dev",
 249                 .mode           = 0555,
 250                 .child          = dev_table,
 251         },
 252 /*
 253  * NOTE: do not add new entries to this table unless you have read
 254  * Documentation/sysctl/ctl_unnumbered.txt
 255  */
 256         { }
 257 };
 258
 259 #ifdef CONFIG_SCHED_DEBUG
 260 static int min_sched_granularity_ns = 100000;           /* 100 usecs */
 261 static int max_sched_granularity_ns = NSEC_PER_SEC;     /* 1 second */
 262 static int min_wakeup_granularity_ns;                   /* 0 usecs */
 263 static int max_wakeup_granularity_ns = NSEC_PER_SEC;    /* 1 second */
 264 static int min_sched_tunable_scaling = SCHED_TUNABLESCALING_NONE;
 265 static int max_sched_tunable_scaling = SCHED_TUNABLESCALING_END-1;
 266 static int min_sched_shares_ratelimit = 100000; /* 100 usec */
 267 static int max_sched_shares_ratelimit = NSEC_PER_SEC; /* 1 second */
 268 #endif
 269
 270 #ifdef CONFIG_COMPACTION
 271 static int min_extfrag_threshold;
 272 static int max_extfrag_threshold = 1000;
 273 #endif
 274
 275 static struct ctl_table kern_table[] = {
 276         {
 277                 .procname       = "sched_child_runs_first",
 278                 .data           = &sysctl_sched_child_runs_first,
 279                 .maxlen         = sizeof(unsigned int),
 280                 .mode           = 0644,
 281                 .proc_handler   = proc_dointvec,
 282         },
 283 #ifdef CONFIG_SCHED_DEBUG
 284         {
 285                 .procname       = "sched_min_granularity_ns",
 286                 .data           = &sysctl_sched_min_granularity,
 287                 .maxlen         = sizeof(unsigned int),
 288                 .mode           = 0644,
 289                 .proc_handler   = sched_proc_update_handler,
 290                 .extra1         = &min_sched_granularity_ns,
 291                 .extra2         = &max_sched_granularity_ns,
 292         },
 293         {
 294                 .procname       = "sched_latency_ns",
 295                 .data           = &sysctl_sched_latency,
 296                 .maxlen         = sizeof(unsigned int),
 297                 .mode           = 0644,
 298                 .proc_handler   = sched_proc_update_handler,
 299                 .extra1         = &min_sched_granularity_ns,
 300                 .extra2         = &max_sched_granularity_ns,
 301         },
 302         {
 303                 .procname       = "sched_wakeup_granularity_ns",
 304                 .data           = &sysctl_sched_wakeup_granularity,
 305                 .maxlen         = sizeof(unsigned int),
 306                 .mode           = 0644,
 307                 .proc_handler   = sched_proc_update_handler,
 308                 .extra1         = &min_wakeup_granularity_ns,
 309                 .extra2         = &max_wakeup_granularity_ns,
 310         },
 311         {
 312                 .procname       = "sched_shares_ratelimit",
 313                 .data           = &sysctl_sched_shares_ratelimit,
 314                 .maxlen         = sizeof(unsigned int),
 315                 .mode           = 0644,
 316                 .proc_handler   = sched_proc_update_handler,
 317                 .extra1         = &min_sched_shares_ratelimit,
 318                 .extra2         = &max_sched_shares_ratelimit,
 319         },
 320         {
 321                 .procname       = "sched_tunable_scaling",
 322                 .data           = &sysctl_sched_tunable_scaling,
 323                 .maxlen         = sizeof(enum sched_tunable_scaling),
 324                 .mode           = 0644,
 325                 .proc_handler   = sched_proc_update_handler,
 326                 .extra1         = &min_sched_tunable_scaling,
 327                 .extra2         = &max_sched_tunable_scaling,
 328         },
 329         {
 330                 .procname       = "sched_shares_thresh",
 331                 .data           = &sysctl_sched_shares_thresh,
 332                 .maxlen         = sizeof(unsigned int),
 333                 .mode           = 0644,
 334                 .proc_handler   = proc_dointvec_minmax,
 335                 .extra1         = &zero,
 336         },
 337         {
 338                 .procname       = "sched_migration_cost",
 339                 .data           = &sysctl_sched_migration_cost,
 340                 .maxlen         = sizeof(unsigned int),
 341                 .mode           = 0644,
 342                 .proc_handler   = proc_dointvec,
 343         },
 344         {
 345                 .procname       = "sched_nr_migrate",
 346                 .data           = &sysctl_sched_nr_migrate,
 347                 .maxlen         = sizeof(unsigned int),
 348                 .mode           = 0644,
 349                 .proc_handler   = proc_dointvec,
 350         },
 351         {
 352                 .procname       = "sched_time_avg",
 353                 .data           = &sysctl_sched_time_avg,
 354                 .maxlen         = sizeof(unsigned int),
 355                 .mode           = 0644,
 356                 .proc_handler   = proc_dointvec,
 357         },
 358         {
 359                 .procname       = "timer_migration",
 360                 .data           = &sysctl_timer_migration,
 361                 .maxlen         = sizeof(unsigned int),
 362                 .mode           = 0644,
 363                 .proc_handler   = proc_dointvec_minmax,
 364                 .extra1         = &zero,
 365                 .extra2         = &one,
 366         },
 367 #endif
 368         {
 369                 .procname       = "sched_rt_period_us",
 370                 .data           = &sysctl_sched_rt_period,
 371                 .maxlen         = sizeof(unsigned int),
 372                 .mode           = 0644,
 373                 .proc_handler   = sched_rt_handler,
 374         },
 375         {
 376                 .procname       = "sched_rt_runtime_us",
 377                 .data           = &sysctl_sched_rt_runtime,
 378                 .maxlen         = sizeof(int),
 379                 .mode           = 0644,
 380                 .proc_handler   = sched_rt_handler,
 381         },
 382         {
 383                 .procname       = "sched_compat_yield",
 384                 .data           = &sysctl_sched_compat_yield,
 385                 .maxlen         = sizeof(unsigned int),
 386                 .mode           = 0644,
 387                 .proc_handler   = proc_dointvec,
 388         },
 389 #ifdef CONFIG_PROVE_LOCKING
 390         {
 391                 .procname       = "prove_locking",
 392                 .data           = &prove_locking,
 393                 .maxlen         = sizeof(int),
 394                 .mode           = 0644,
 395                 .proc_handler   = proc_dointvec,
 396         },
 397 #endif
 398 #ifdef CONFIG_LOCK_STAT
 399         {
 400                 .procname       = "lock_stat",
 401                 .data           = &lock_stat,
 402                 .maxlen         = sizeof(int),
 403                 .mode           = 0644,
 404                 .proc_handler   = proc_dointvec,
 405         },
 406 #endif
 407         {
 408                 .procname       = "panic",
 409                 .data           = &panic_timeout,
 410                 .maxlen         = sizeof(int),
 411                 .mode           = 0644,
 412                 .proc_handler   = proc_dointvec,
 413         },
 414         {
 415                 .procname       = "core_uses_pid",
 416                 .data           = &core_uses_pid,
 417                 .maxlen         = sizeof(int),
 418                 .mode           = 0644,
 419                 .proc_handler   = proc_dointvec,
 420         },
 421         {
 422                 .procname       = "core_pattern",
 423                 .data           = core_pattern,
 424                 .maxlen         = CORENAME_MAX_SIZE,
 425                 .mode           = 0644,
 426                 .proc_handler   = proc_dostring,
 427         },
 428         {
 429                 .procname       = "core_pipe_limit",
 430                 .data           = &core_pipe_limit,
 431                 .maxlen         = sizeof(unsigned int),
 432                 .mode           = 0644,
 433                 .proc_handler   = proc_dointvec,
 434         },
 435 #ifdef CONFIG_PROC_SYSCTL
 436         {
 437                 .procname       = "tainted",
 438                 .maxlen         = sizeof(long),
 439                 .mode           = 0644,
 440                 .proc_handler   = proc_taint,
 441         },
 442 #endif
 443 #ifdef CONFIG_LATENCYTOP
 444         {
 445                 .procname       = "latencytop",
 446                 .data           = &latencytop_enabled,
 447                 .maxlen         = sizeof(int),
 448                 .mode           = 0644,
 449                 .proc_handler   = proc_dointvec,
 450         },
 451 #endif
 452 #ifdef CONFIG_BLK_DEV_INITRD
 453         {
 454                 .procname       = "real-root-dev",
 455                 .data           = &real_root_dev,
 456                 .maxlen         = sizeof(int),
 457                 .mode           = 0644,
 458                 .proc_handler   = proc_dointvec,
 459         },
 460 #endif
 461         {
 462                 .procname       = "print-fatal-signals",
 463                 .data           = &print_fatal_signals,
 464                 .maxlen         = sizeof(int),
 465                 .mode           = 0644,
 466                 .proc_handler   = proc_dointvec,
 467         },
 468 #ifdef CONFIG_SPARC
 469         {
 470                 .procname       = "reboot-cmd",
 471                 .data           = reboot_command,
 472                 .maxlen         = 256,
 473                 .mode           = 0644,
 474                 .proc_handler   = proc_dostring,
 475         },
 476         {
 477                 .procname       = "stop-a",
 478                 .data           = &stop_a_enabled,
 479                 .maxlen         = sizeof (int),
 480                 .mode           = 0644,
 481                 .proc_handler   = proc_dointvec,
 482         },
 483         {
 484                 .procname       = "scons-poweroff",
 485                 .data           = &scons_pwroff,
 486                 .maxlen         = sizeof (int),
 487                 .mode           = 0644,
 488                 .proc_handler   = proc_dointvec,
 489         },
 490 #endif
 491 #ifdef CONFIG_SPARC64
 492         {
 493                 .procname       = "tsb-ratio",
 494                 .data           = &sysctl_tsb_ratio,
 495                 .maxlen         = sizeof (int),
 496                 .mode           = 0644,
 497                 .proc_handler   = proc_dointvec,
 498         },
 499 #endif
 500 #ifdef __hppa__
 501         {
 502                 .procname       = "soft-power",
 503                 .data           = &pwrsw_enabled,
 504                 .maxlen         = sizeof (int),
 505                 .mode           = 0644,
 506                 .proc_handler   = proc_dointvec,
 507         },
 508         {
 509                 .procname       = "unaligned-trap",
 510                 .data           = &unaligned_enabled,
 511                 .maxlen         = sizeof (int),
 512                 .mode           = 0644,
 513                 .proc_handler   = proc_dointvec,
 514         },
 515 #endif
 516         {
 517                 .procname       = "ctrl-alt-del",
 518                 .data           = &C_A_D,
 519                 .maxlen         = sizeof(int),
 520                 .mode           = 0644,
 521                 .proc_handler   = proc_dointvec,
 522         },
 523 #ifdef CONFIG_FUNCTION_TRACER
 524         {
 525                 .procname       = "ftrace_enabled",
 526                 .data           = &ftrace_enabled,
 527                 .maxlen         = sizeof(int),
 528                 .mode           = 0644,
 529                 .proc_handler   = ftrace_enable_sysctl,
 530         },
 531 #endif
 532 #ifdef CONFIG_STACK_TRACER
 533         {
 534                 .procname       = "stack_tracer_enabled",
 535                 .data           = &stack_tracer_enabled,
 536                 .maxlen         = sizeof(int),
 537                 .mode           = 0644,
 538                 .proc_handler   = stack_trace_sysctl,
 539         },
 540 #endif
 541 #ifdef CONFIG_TRACING
 542         {
 543                 .procname       = "ftrace_dump_on_oops",
 544                 .data           = &ftrace_dump_on_oops,
 545                 .maxlen         = sizeof(int),
 546                 .mode           = 0644,
 547                 .proc_handler   = proc_dointvec,
 548         },
 549 #endif
 550 #ifdef CONFIG_MODULES
 551         {
 552                 .procname       = "modprobe",
 553                 .data           = &modprobe_path,
 554                 .maxlen         = KMOD_PATH_LEN,
 555                 .mode           = 0644,
 556                 .proc_handler   = proc_dostring,
 557         },
 558         {
 559                 .procname       = "modules_disabled",
 560                 .data           = &modules_disabled,
 561                 .maxlen         = sizeof(int),
 562                 .mode           = 0644,
 563                 /* only handle a transition from default "0" to "1" */
 564                 .proc_handler   = proc_dointvec_minmax,
 565                 .extra1         = &one,
 566                 .extra2         = &one,
 567         },
 568 #endif
 569 #ifdef CONFIG_HOTPLUG
 570         {
 571                 .procname       = "hotplug",
 572                 .data           = &uevent_helper,
 573                 .maxlen         = UEVENT_HELPER_PATH_LEN,
 574                 .mode           = 0644,
 575                 .proc_handler   = proc_dostring,
 576         },
 577 #endif
 578 #ifdef CONFIG_CHR_DEV_SG
 579         {
 580                 .procname       = "sg-big-buff",
 581                 .data           = &sg_big_buff,
 582                 .maxlen         = sizeof (int),
 583                 .mode           = 0444,
 584                 .proc_handler   = proc_dointvec,
 585         },
 586 #endif
 587 #ifdef CONFIG_BSD_PROCESS_ACCT
 588         {
 589                 .procname       = "acct",
 590                 .data           = &acct_parm,
 591                 .maxlen         = 3*sizeof(int),
 592                 .mode           = 0644,
 593                 .proc_handler   = proc_dointvec,
 594         },
 595 #endif
 596 #ifdef CONFIG_MAGIC_SYSRQ
 597         {
 598                 .procname       = "sysrq",
 599                 .data           = &__sysrq_enabled,
 600                 .maxlen         = sizeof (int),
 601                 .mode           = 0644,
 602                 .proc_handler   = sysrq_sysctl_handler,
 603         },
 604 #endif
 605 #ifdef CONFIG_PROC_SYSCTL
 606         {
 607                 .procname       = "cad_pid",
 608                 .data           = NULL,
 609                 .maxlen         = sizeof (int),
 610                 .mode           = 0600,
 611                 .proc_handler   = proc_do_cad_pid,
 612         },
 613 #endif
 614         {
 615                 .procname       = "threads-max",
 616                 .data           = &max_threads,
 617                 .maxlen         = sizeof(int),
 618                 .mode           = 0644,
 619                 .proc_handler   = proc_dointvec,
 620         },
 621         {
 622                 .procname       = "random",
 623                 .mode           = 0555,
 624                 .child          = random_table,
 625         },
 626         {
 627                 .procname       = "overflowuid",
 628                 .data           = &overflowuid,
 629                 .maxlen         = sizeof(int),
 630                 .mode           = 0644,
 631                 .proc_handler   = proc_dointvec_minmax,
 632                 .extra1         = &minolduid,
 633                 .extra2         = &maxolduid,
 634         },
 635         {
 636                 .procname       = "overflowgid",
 637                 .data           = &overflowgid,
 638                 .maxlen         = sizeof(int),
 639                 .mode           = 0644,
 640                 .proc_handler   = proc_dointvec_minmax,
 641                 .extra1         = &minolduid,
 642                 .extra2         = &maxolduid,
 643         },
 644 #ifdef CONFIG_S390
 645 #ifdef CONFIG_MATHEMU
 646         {
 647                 .procname       = "ieee_emulation_warnings",
 648                 .data           = &sysctl_ieee_emulation_warnings,
 649                 .maxlen         = sizeof(int),
 650                 .mode           = 0644,
 651                 .proc_handler   = proc_dointvec,
 652         },
 653 #endif
 654         {
 655                 .procname       = "userprocess_debug",
 656                 .data           = &show_unhandled_signals,
 657                 .maxlen         = sizeof(int),
 658                 .mode           = 0644,
 659                 .proc_handler   = proc_dointvec,
 660         },
 661 #endif
 662         {
 663                 .procname       = "pid_max",
 664                 .data           = &pid_max,
 665                 .maxlen         = sizeof (int),
 666                 .mode           = 0644,
 667                 .proc_handler   = proc_dointvec_minmax,
 668                 .extra1         = &pid_max_min,
 669                 .extra2         = &pid_max_max,
 670         },
 671         {
 672                 .procname       = "panic_on_oops",
 673                 .data           = &panic_on_oops,
 674                 .maxlen         = sizeof(int),
 675                 .mode           = 0644,
 676                 .proc_handler   = proc_dointvec,
 677         },
 678 #if defined CONFIG_PRINTK
 679         {
 680                 .procname       = "printk",
 681                 .data           = &console_loglevel,
 682                 .maxlen         = 4*sizeof(int),
 683                 .mode           = 0644,
 684                 .proc_handler   = proc_dointvec,
 685         },
 686         {
 687                 .procname       = "printk_ratelimit",
 688                 .data           = &printk_ratelimit_state.interval,
 689                 .maxlen         = sizeof(int),
 690                 .mode           = 0644,
 691                 .proc_handler   = proc_dointvec_jiffies,
 692         },
 693         {
 694                 .procname       = "printk_ratelimit_burst",
 695                 .data           = &printk_ratelimit_state.burst,
 696                 .maxlen         = sizeof(int),
 697                 .mode           = 0644,
 698                 .proc_handler   = proc_dointvec,
 699         },
 700         {
 701                 .procname       = "printk_delay",
 702                 .data           = &printk_delay_msec,
 703                 .maxlen         = sizeof(int),
 704                 .mode           = 0644,
 705                 .proc_handler   = proc_dointvec_minmax,
 706                 .extra1         = &zero,
 707                 .extra2         = &ten_thousand,
 708         },
 709 #endif
 710         {
 711                 .procname       = "ngroups_max",
 712                 .data           = &ngroups_max,
 713                 .maxlen         = sizeof (int),
 714                 .mode           = 0444,
 715                 .proc_handler   = proc_dointvec,
 716         },
 717 #if defined(CONFIG_LOCKUP_DETECTOR)
 718         {
 719                 .procname       = "watchdog",
 720                 .data           = &watchdog_enabled,
 721                 .maxlen         = sizeof (int),
 722                 .mode           = 0644,
 723                 .proc_handler   = proc_dowatchdog_enabled,
 724         },
 725         {
 726                 .procname       = "watchdog_thresh",
 727                 .data           = &softlockup_thresh,
 728                 .maxlen         = sizeof(int),
 729                 .mode           = 0644,
 730                 .proc_handler   = proc_dowatchdog_thresh,
 731                 .extra1         = &neg_one,
 732                 .extra2         = &sixty,
 733         },
 734         {
 735                 .procname       = "softlockup_panic",
 736                 .data           = &softlockup_panic,
 737                 .maxlen         = sizeof(int),
 738                 .mode           = 0644,
 739                 .proc_handler   = proc_dointvec_minmax,
 740                 .extra1         = &zero,
 741                 .extra2         = &one,
 742         },
 743 #endif
 744 #if defined(CONFIG_X86_LOCAL_APIC) && defined(CONFIG_X86) && !defined(CONFIG_LOCKUP_DETECTOR)
 745         {
 746                 .procname       = "unknown_nmi_panic",
 747                 .data           = &unknown_nmi_panic,
 748                 .maxlen         = sizeof (int),
 749                 .mode           = 0644,
 750                 .proc_handler   = proc_dointvec,
 751         },
 752         {
 753                 .procname       = "nmi_watchdog",
 754                 .data           = &nmi_watchdog_enabled,
 755                 .maxlen         = sizeof (int),
 756                 .mode           = 0644,
 757                 .proc_handler   = proc_nmi_enabled,
 758         },
 759 #endif
 760 #if defined(CONFIG_X86)
 761         {
 762                 .procname       = "panic_on_unrecovered_nmi",
 763                 .data           = &panic_on_unrecovered_nmi,
 764                 .maxlen         = sizeof(int),
 765                 .mode           = 0644,
 766                 .proc_handler   = proc_dointvec,
 767         },
 768         {
 769                 .procname       = "panic_on_io_nmi",
 770                 .data           = &panic_on_io_nmi,
 771                 .maxlen         = sizeof(int),
 772                 .mode           = 0644,
 773                 .proc_handler   = proc_dointvec,
 774         },
 775         {
 776                 .procname       = "bootloader_type",
 777                 .data           = &bootloader_type,
 778                 .maxlen         = sizeof (int),
 779                 .mode           = 0444,
 780                 .proc_handler   = proc_dointvec,
 781         },
 782         {
 783                 .procname       = "bootloader_version",
 784                 .data           = &bootloader_version,
 785                 .maxlen         = sizeof (int),
 786                 .mode           = 0444,
 787                 .proc_handler   = proc_dointvec,
 788         },
 789         {
 790                 .procname       = "kstack_depth_to_print",
 791                 .data           = &kstack_depth_to_print,
 792                 .maxlen         = sizeof(int),
 793                 .mode           = 0644,
 794                 .proc_handler   = proc_dointvec,
 795         },
 796         {
 797                 .procname       = "io_delay_type",
 798                 .data           = &io_delay_type,
 799                 .maxlen         = sizeof(int),
 800                 .mode           = 0644,
 801                 .proc_handler   = proc_dointvec,
 802         },
 803 #endif
 804 #if defined(CONFIG_MMU)
 805         {
 806                 .procname       = "randomize_va_space",
 807                 .data           = &randomize_va_space,
 808                 .maxlen         = sizeof(int),
 809                 .mode           = 0644,
 810                 .proc_handler   = proc_dointvec,
 811         },
 812 #endif
 813 #if defined(CONFIG_S390) && defined(CONFIG_SMP)
 814         {
 815                 .procname       = "spin_retry",
 816                 .data           = &spin_retry,
 817                 .maxlen         = sizeof (int),
 818                 .mode           = 0644,
 819                 .proc_handler   = proc_dointvec,
 820         },
 821 #endif
 822 #if     defined(CONFIG_ACPI_SLEEP) && defined(CONFIG_X86)
 823         {
 824                 .procname       = "acpi_video_flags",
 825                 .data           = &acpi_realmode_flags,
 826                 .maxlen         = sizeof (unsigned long),
 827                 .mode           = 0644,
 828                 .proc_handler   = proc_doulongvec_minmax,
 829         },
 830 #endif
 831 #ifdef CONFIG_IA64
 832         {
 833                 .procname       = "ignore-unaligned-usertrap",
 834                 .data           = &no_unaligned_warning,
 835                 .maxlen         = sizeof (int),
 836                 .mode           = 0644,
 837                 .proc_handler   = proc_dointvec,
 838         },
 839         {
 840                 .procname       = "unaligned-dump-stack",
 841                 .data           = &unaligned_dump_stack,
 842                 .maxlen         = sizeof (int),
 843                 .mode           = 0644,
 844                 .proc_handler   = proc_dointvec,
 845         },
 846 #endif
 847 #ifdef CONFIG_DETECT_HUNG_TASK
 848         {
 849                 .procname       = "hung_task_panic",
 850                 .data           = &sysctl_hung_task_panic,
 851                 .maxlen         = sizeof(int),
 852                 .mode           = 0644,
 853                 .proc_handler   = proc_dointvec_minmax,
 854                 .extra1         = &zero,
 855                 .extra2         = &one,
 856         },
 857         {
 858                 .procname       = "hung_task_check_count",
 859                 .data           = &sysctl_hung_task_check_count,
 860                 .maxlen         = sizeof(unsigned long),
 861                 .mode           = 0644,
 862                 .proc_handler   = proc_doulongvec_minmax,
 863         },
 864         {
 865                 .procname       = "hung_task_timeout_secs",
 866                 .data           = &sysctl_hung_task_timeout_secs,
 867                 .maxlen         = sizeof(unsigned long),
 868                 .mode           = 0644,
 869                 .proc_handler   = proc_dohung_task_timeout_secs,
 870         },
 871         {
 872                 .procname       = "hung_task_warnings",
 873                 .data           = &sysctl_hung_task_warnings,
 874                 .maxlen         = sizeof(unsigned long),
 875                 .mode           = 0644,
 876                 .proc_handler   = proc_doulongvec_minmax,
 877         },
 878 #endif
 879 #ifdef CONFIG_COMPAT
 880         {
 881                 .procname       = "compat-log",
 882                 .data           = &compat_log,
 883                 .maxlen         = sizeof (int),
 884                 .mode           = 0644,
 885                 .proc_handler   = proc_dointvec,
 886         },
 887 #endif
 888 #ifdef CONFIG_RT_MUTEXES
 889         {
 890                 .procname       = "max_lock_depth",
 891                 .data           = &max_lock_depth,
 892                 .maxlen         = sizeof(int),
 893                 .mode           = 0644,
 894                 .proc_handler   = proc_dointvec,
 895         },
 896 #endif
 897         {
 898                 .procname       = "poweroff_cmd",
 899                 .data           = &poweroff_cmd,
 900                 .maxlen         = POWEROFF_CMD_PATH_LEN,
 901                 .mode           = 0644,
 902                 .proc_handler   = proc_dostring,
 903         },
 904 #ifdef CONFIG_KEYS
 905         {
 906                 .procname       = "keys",
 907                 .mode           = 0555,
 908                 .child          = key_sysctls,
 909         },
 910 #endif
 911 #ifdef CONFIG_RCU_TORTURE_TEST
 912         {
 913                 .procname       = "rcutorture_runnable",
 914                 .data           = &rcutorture_runnable,
 915                 .maxlen         = sizeof(int),
 916                 .mode           = 0644,
 917                 .proc_handler   = proc_dointvec,
 918         },
 919 #endif
 920 #ifdef CONFIG_SLOW_WORK
 921         {
 922                 .procname       = "slow-work",
 923                 .mode           = 0555,
 924                 .child          = slow_work_sysctls,
 925         },
 926 #endif
 927 #ifdef CONFIG_PERF_EVENTS
 928         {
 929                 .procname       = "perf_event_paranoid",
 930                 .data           = &sysctl_perf_event_paranoid,
 931                 .maxlen         = sizeof(sysctl_perf_event_paranoid),
 932                 .mode           = 0644,
 933                 .proc_handler   = proc_dointvec,
 934         },
 935         {
 936                 .procname       = "perf_event_mlock_kb",
 937                 .data           = &sysctl_perf_event_mlock,
 938                 .maxlen         = sizeof(sysctl_perf_event_mlock),
 939                 .mode           = 0644,
 940                 .proc_handler   = proc_dointvec,
 941         },
 942         {
 943                 .procname       = "perf_event_max_sample_rate",
 944                 .data           = &sysctl_perf_event_sample_rate,
 945                 .maxlen         = sizeof(sysctl_perf_event_sample_rate),
 946                 .mode           = 0644,
 947                 .proc_handler   = proc_dointvec,
 948         },
 949 #endif
 950 #ifdef CONFIG_KMEMCHECK
 951         {
 952                 .procname       = "kmemcheck",
 953                 .data           = &kmemcheck_enabled,
 954                 .maxlen         = sizeof(int),
 955                 .mode           = 0644,
 956                 .proc_handler   = proc_dointvec,
 957         },
 958 #endif
 959 #ifdef CONFIG_BLOCK
 960         {
 961                 .procname       = "blk_iopoll",
 962                 .data           = &blk_iopoll_enabled,
 963                 .maxlen         = sizeof(int),
 964                 .mode           = 0644,
 965                 .proc_handler   = proc_dointvec,
 966         },
 967 #endif
 968 /*
 969  * NOTE: do not add new entries to this table unless you have read
 970  * Documentation/sysctl/ctl_unnumbered.txt
 971  */
 972         { }
 973 };
 974
 975 static struct ctl_table vm_table[] = {
 976         {
 977                 .procname       = "overcommit_memory",
 978                 .data           = &sysctl_overcommit_memory,
 979                 .maxlen         = sizeof(sysctl_overcommit_memory),
 980                 .mode           = 0644,
 981                 .proc_handler   = proc_dointvec,
 982         },
 983         {
 984                 .procname       = "panic_on_oom",
 985                 .data           = &sysctl_panic_on_oom,
 986                 .maxlen         = sizeof(sysctl_panic_on_oom),
 987                 .mode           = 0644,
 988                 .proc_handler   = proc_dointvec,
 989         },
 990         {
 991                 .procname       = "oom_kill_allocating_task",
 992                 .data           = &sysctl_oom_kill_allocating_task,
 993                 .maxlen         = sizeof(sysctl_oom_kill_allocating_task),
 994                 .mode           = 0644,
 995                 .proc_handler   = proc_dointvec,
 996         },
 997         {
 998                 .procname       = "oom_dump_tasks",
 999                 .data           = &sysctl_oom_dump_tasks,
1000                 .maxlen         = sizeof(sysctl_oom_dump_tasks),
1001                 .mode           = 0644,
1002                 .proc_handler   = proc_dointvec,
1003         },
1004         {
1005                 .procname       = "overcommit_ratio",
1006                 .data           = &sysctl_overcommit_ratio,
1007                 .maxlen         = sizeof(sysctl_overcommit_ratio),
1008                 .mode           = 0644,
1009                 .proc_handler   = proc_dointvec,
1010         },
1011         {
1012                 .procname       = "page-cluster",
1013                 .data           = &page_cluster,
1014                 .maxlen         = sizeof(int),
1015                 .mode           = 0644,
1016                 .proc_handler   = proc_dointvec,
1017         },
1018         {
1019                 .procname       = "dirty_background_ratio",
1020                 .data           = &dirty_background_ratio,
1021                 .maxlen         = sizeof(dirty_background_ratio),
1022                 .mode           = 0644,
1023                 .proc_handler   = dirty_background_ratio_handler,
1024                 .extra1         = &zero,
1025                 .extra2         = &one_hundred,
1026         },
1027         {
1028                 .procname       = "dirty_background_bytes",
1029                 .data           = &dirty_background_bytes,
1030                 .maxlen         = sizeof(dirty_background_bytes),
1031                 .mode           = 0644,
1032                 .proc_handler   = dirty_background_bytes_handler,
1033                 .extra1         = &one_ul,
1034         },
1035         {
1036                 .procname       = "dirty_ratio",
1037                 .data           = &vm_dirty_ratio,
1038                 .maxlen         = sizeof(vm_dirty_ratio),
1039                 .mode           = 0644,
1040                 .proc_handler   = dirty_ratio_handler,
1041                 .extra1         = &zero,
1042                 .extra2         = &one_hundred,
1043         },
1044         {
1045                 .procname       = "dirty_bytes",
1046                 .data           = &vm_dirty_bytes,
1047                 .maxlen         = sizeof(vm_dirty_bytes),
1048                 .mode           = 0644,
1049                 .proc_handler   = dirty_bytes_handler,
1050                 .extra1         = &dirty_bytes_min,
1051         },
1052         {
1053                 .procname       = "dirty_writeback_centisecs",
1054                 .data           = &dirty_writeback_interval,
1055                 .maxlen         = sizeof(dirty_writeback_interval),
1056                 .mode           = 0644,
1057                 .proc_handler   = dirty_writeback_centisecs_handler,
1058         },
1059         {
1060                 .procname       = "dirty_expire_centisecs",
1061                 .data           = &dirty_expire_interval,
1062                 .maxlen         = sizeof(dirty_expire_interval),
1063                 .mode           = 0644,
1064                 .proc_handler   = proc_dointvec,
1065         },
1066         {
1067                 .procname       = "nr_pdflush_threads",
1068                 .data           = &nr_pdflush_threads,
1069                 .maxlen         = sizeof nr_pdflush_threads,
1070                 .mode           = 0444 /* read-only*/,
1071                 .proc_handler   = proc_dointvec,
1072         },
1073         {
1074                 .procname       = "swappiness",
1075                 .data           = &vm_swappiness,
1076                 .maxlen         = sizeof(vm_swappiness),
1077                 .mode           = 0644,
1078                 .proc_handler   = proc_dointvec_minmax,
1079                 .extra1         = &zero,
1080                 .extra2         = &one_hundred,
1081         },
1082 #ifdef CONFIG_HUGETLB_PAGE
1083         {
1084                 .procname       = "nr_hugepages",
1085                 .data           = NULL,
1086                 .maxlen         = sizeof(unsigned long),
1087                 .mode           = 0644,
1088                 .proc_handler   = hugetlb_sysctl_handler,
1089                 .extra1         = (void *)&hugetlb_zero,
1090                 .extra2         = (void *)&hugetlb_infinity,
1091         },
1092 #ifdef CONFIG_NUMA
1093         {
1094                 .procname       = "nr_hugepages_mempolicy",
1095                 .data           = NULL,
1096                 .maxlen         = sizeof(unsigned long),
1097                 .mode           = 0644,
1098                 .proc_handler   = &hugetlb_mempolicy_sysctl_handler,
1099                 .extra1         = (void *)&hugetlb_zero,
1100                 .extra2         = (void *)&hugetlb_infinity,
1101         },
1102 #endif
1103          {
1104                 .procname       = "hugetlb_shm_group",
1105                 .data           = &sysctl_hugetlb_shm_group,
1106                 .maxlen         = sizeof(gid_t),
1107                 .mode           = 0644,
1108                 .proc_handler   = proc_dointvec,
1109          },
1110          {
1111                 .procname       = "hugepages_treat_as_movable",
1112                 .data           = &hugepages_treat_as_movable,
1113                 .maxlen         = sizeof(int),
1114                 .mode           = 0644,
1115                 .proc_handler   = hugetlb_treat_movable_handler,
1116         },
1117         {
1118                 .procname       = "nr_overcommit_hugepages",
1119                 .data           = NULL,
1120                 .maxlen         = sizeof(unsigned long),
1121                 .mode           = 0644,
1122                 .proc_handler   = hugetlb_overcommit_handler,
1123                 .extra1         = (void *)&hugetlb_zero,
1124                 .extra2         = (void *)&hugetlb_infinity,
1125         },
1126 #endif
1127         {
1128                 .procname       = "lowmem_reserve_ratio",
1129                 .data           = &sysctl_lowmem_reserve_ratio,
1130                 .maxlen         = sizeof(sysctl_lowmem_reserve_ratio),
1131                 .mode           = 0644,
1132                 .proc_handler   = lowmem_reserve_ratio_sysctl_handler,
1133         },
1134         {
1135                 .procname       = "drop_caches",
1136                 .data           = &sysctl_drop_caches,
1137                 .maxlen         = sizeof(int),
1138                 .mode           = 0644,
1139                 .proc_handler   = drop_caches_sysctl_handler,
1140         },
1141 #ifdef CONFIG_COMPACTION
1142         {
1143                 .procname       = "compact_memory",
1144                 .data           = &sysctl_compact_memory,
1145                 .maxlen         = sizeof(int),
1146                 .mode           = 0200,
1147                 .proc_handler   = sysctl_compaction_handler,
1148         },
1149         {
1150                 .procname       = "extfrag_threshold",
1151                 .data           = &sysctl_extfrag_threshold,
1152                 .maxlen         = sizeof(int),
1153                 .mode           = 0644,
1154                 .proc_handler   = sysctl_extfrag_handler,
1155                 .extra1         = &min_extfrag_threshold,
1156                 .extra2         = &max_extfrag_threshold,
1157         },
1158
1159 #endif /* CONFIG_COMPACTION */
1160         {
1161                 .procname       = "min_free_kbytes",
1162                 .data           = &min_free_kbytes,
1163                 .maxlen         = sizeof(min_free_kbytes),
1164                 .mode           = 0644,
1165                 .proc_handler   = min_free_kbytes_sysctl_handler,
1166                 .extra1         = &zero,
1167         },
1168         {
1169                 .procname       = "percpu_pagelist_fraction",
1170                 .data           = &percpu_pagelist_fraction,
1171                 .maxlen         = sizeof(percpu_pagelist_fraction),
1172                 .mode           = 0644,
1173                 .proc_handler   = percpu_pagelist_fraction_sysctl_handler,
1174                 .extra1         = &min_percpu_pagelist_fract,
1175         },
1176 #ifdef CONFIG_MMU
1177         {
1178                 .procname       = "max_map_count",
1179                 .data           = &sysctl_max_map_count,
1180                 .maxlen         = sizeof(sysctl_max_map_count),
1181                 .mode           = 0644,
1182                 .proc_handler   = proc_dointvec_minmax,
1183                 .extra1         = &zero,
1184         },
1185 #else
1186         {
1187                 .procname       = "nr_trim_pages",
1188                 .data           = &sysctl_nr_trim_pages,
1189                 .maxlen         = sizeof(sysctl_nr_trim_pages),
1190                 .mode           = 0644,
1191                 .proc_handler   = proc_dointvec_minmax,
1192                 .extra1         = &zero,
1193         },
1194 #endif
1195         {
1196                 .procname       = "laptop_mode",
1197                 .data           = &laptop_mode,
1198                 .maxlen         = sizeof(laptop_mode),
1199                 .mode           = 0644,
1200                 .proc_handler   = proc_dointvec_jiffies,
1201         },
1202         {
1203                 .procname       = "block_dump",
1204                 .data           = &block_dump,
1205                 .maxlen         = sizeof(block_dump),
1206                 .mode           = 0644,
1207                 .proc_handler   = proc_dointvec,
1208                 .extra1         = &zero,
1209         },
1210         {
1211                 .procname       = "vfs_cache_pressure",
1212                 .data           = &sysctl_vfs_cache_pressure,
1213                 .maxlen         = sizeof(sysctl_vfs_cache_pressure),
1214                 .mode           = 0644,
1215                 .proc_handler   = proc_dointvec,
1216                 .extra1         = &zero,
1217         },
1218 #ifdef HAVE_ARCH_PICK_MMAP_LAYOUT
1219         {
1220                 .procname       = "legacy_va_layout",
1221                 .data           = &sysctl_legacy_va_layout,
1222                 .maxlen         = sizeof(sysctl_legacy_va_layout),
1223                 .mode           = 0644,
1224                 .proc_handler   = proc_dointvec,
1225                 .extra1         = &zero,
1226         },
1227 #endif
1228 #ifdef CONFIG_NUMA
1229         {
1230                 .procname       = "zone_reclaim_mode",
1231                 .data           = &zone_reclaim_mode,
1232                 .maxlen         = sizeof(zone_reclaim_mode),
1233                 .mode           = 0644,
1234                 .proc_handler   = proc_dointvec,
1235                 .extra1         = &zero,
1236         },
1237         {
1238                 .procname       = "min_unmapped_ratio",
1239                 .data           = &sysctl_min_unmapped_ratio,
1240                 .maxlen         = sizeof(sysctl_min_unmapped_ratio),
1241                 .mode           = 0644,
1242                 .proc_handler   = sysctl_min_unmapped_ratio_sysctl_handler,
1243                 .extra1         = &zero,
1244                 .extra2         = &one_hundred,
1245         },
1246         {
1247                 .procname       = "min_slab_ratio",
1248                 .data           = &sysctl_min_slab_ratio,
1249                 .maxlen         = sizeof(sysctl_min_slab_ratio),
1250                 .mode           = 0644,
1251                 .proc_handler   = sysctl_min_slab_ratio_sysctl_handler,
1252                 .extra1         = &zero,
1253                 .extra2         = &one_hundred,
1254         },
1255 #endif
1256 #ifdef CONFIG_SMP
1257         {
1258                 .procname       = "stat_interval",
1259                 .data           = &sysctl_stat_interval,
1260                 .maxlen         = sizeof(sysctl_stat_interval),
1261                 .mode           = 0644,
1262                 .proc_handler   = proc_dointvec_jiffies,
1263         },
1264 #endif
1265 #ifdef CONFIG_MMU
1266         {
1267                 .procname       = "mmap_min_addr",
1268                 .data           = &dac_mmap_min_addr,
1269                 .maxlen         = sizeof(unsigned long),
1270                 .mode           = 0644,
1271                 .proc_handler   = mmap_min_addr_handler,
1272         },
1273 #endif
1274 #ifdef CONFIG_NUMA
1275         {
1276                 .procname       = "numa_zonelist_order",
1277                 .data           = &numa_zonelist_order,
1278                 .maxlen         = NUMA_ZONELIST_ORDER_LEN,
1279                 .mode           = 0644,
1280                 .proc_handler   = numa_zonelist_order_handler,
1281         },
1282 #endif
1283 #if (defined(CONFIG_X86_32) && !defined(CONFIG_UML))|| \
1284    (defined(CONFIG_SUPERH) && defined(CONFIG_VSYSCALL))
1285         {
1286                 .procname       = "vdso_enabled",
1287                 .data           = &vdso_enabled,
1288                 .maxlen         = sizeof(vdso_enabled),
1289                 .mode           = 0644,
1290                 .proc_handler   = proc_dointvec,
1291                 .extra1         = &zero,
1292         },
1293 #endif
1294 #ifdef CONFIG_HIGHMEM
1295         {
1296                 .procname       = "highmem_is_dirtyable",
1297                 .data           = &vm_highmem_is_dirtyable,
1298                 .maxlen         = sizeof(vm_highmem_is_dirtyable),
1299                 .mode           = 0644,
1300                 .proc_handler   = proc_dointvec_minmax,
1301                 .extra1         = &zero,
1302                 .extra2         = &one,
1303         },
1304 #endif
1305         {
1306                 .procname       = "scan_unevictable_pages",
1307                 .data           = &scan_unevictable_pages,
1308                 .maxlen         = sizeof(scan_unevictable_pages),
1309                 .mode           = 0644,
1310                 .proc_handler   = scan_unevictable_handler,
1311         },
1312 #ifdef CONFIG_MEMORY_FAILURE
1313         {
1314                 .procname       = "memory_failure_early_kill",
1315                 .data           = &sysctl_memory_failure_early_kill,
1316                 .maxlen         = sizeof(sysctl_memory_failure_early_kill),
1317                 .mode           = 0644,
1318                 .proc_handler   = proc_dointvec_minmax,
1319                 .extra1         = &zero,
1320                 .extra2         = &one,
1321         },
1322         {
1323                 .procname       = "memory_failure_recovery",
1324                 .data           = &sysctl_memory_failure_recovery,
1325                 .maxlen         = sizeof(sysctl_memory_failure_recovery),
1326                 .mode           = 0644,
1327                 .proc_handler   = proc_dointvec_minmax,
1328                 .extra1         = &zero,
1329                 .extra2         = &one,
1330         },
1331 #endif
1332
1333 /*
1334  * NOTE: do not add new entries to this table unless you have read
1335  * Documentation/sysctl/ctl_unnumbered.txt
1336  */
1337         { }
1338 };
1339
1340 #if defined(CONFIG_BINFMT_MISC) || defined(CONFIG_BINFMT_MISC_MODULE)
1341 static struct ctl_table binfmt_misc_table[] = {
1342         { }
1343 };
1344 #endif
1345
1346 static struct ctl_table fs_table[] = {
1347         {
1348                 .procname       = "inode-nr",
1349                 .data           = &inodes_stat,
1350                 .maxlen         = 2*sizeof(int),
1351                 .mode           = 0444,
1352                 .proc_handler   = proc_dointvec,
1353         },
1354         {
1355                 .procname       = "inode-state",
1356                 .data           = &inodes_stat,
1357                 .maxlen         = 7*sizeof(int),
1358                 .mode           = 0444,
1359                 .proc_handler   = proc_dointvec,
1360         },
1361         {
1362                 .procname       = "file-nr",
1363                 .data           = &files_stat,
1364                 .maxlen         = 3*sizeof(int),
1365                 .mode           = 0444,
1366                 .proc_handler   = proc_nr_files,
1367         },
1368         {
1369                 .procname       = "file-max",
1370                 .data           = &files_stat.max_files,
1371                 .maxlen         = sizeof(int),
1372                 .mode           = 0644,
1373                 .proc_handler   = proc_dointvec,
1374         },
1375         {
1376                 .procname       = "nr_open",
1377                 .data           = &sysctl_nr_open,
1378                 .maxlen         = sizeof(int),
1379                 .mode           = 0644,
1380                 .proc_handler   = proc_dointvec_minmax,
1381                 .extra1         = &sysctl_nr_open_min,
1382                 .extra2         = &sysctl_nr_open_max,
1383         },
1384         {
1385                 .procname       = "dentry-state",
1386                 .data           = &dentry_stat,
1387                 .maxlen         = 6*sizeof(int),
1388                 .mode           = 0444,
1389                 .proc_handler   = proc_dointvec,
1390         },
1391         {
1392                 .procname       = "overflowuid",
1393                 .data           = &fs_overflowuid,
1394                 .maxlen         = sizeof(int),
1395                 .mode           = 0644,
1396                 .proc_handler   = proc_dointvec_minmax,
1397                 .extra1         = &minolduid,
1398                 .extra2         = &maxolduid,
1399         },
1400         {
1401                 .procname       = "overflowgid",
1402                 .data           = &fs_overflowgid,
1403                 .maxlen         = sizeof(int),
1404                 .mode           = 0644,
1405                 .proc_handler   = proc_dointvec_minmax,
1406                 .extra1         = &minolduid,
1407                 .extra2         = &maxolduid,
1408         },
1409 #ifdef CONFIG_FILE_LOCKING
1410         {
1411                 .procname       = "leases-enable",
1412                 .data           = &leases_enable,
1413                 .maxlen         = sizeof(int),
1414                 .mode           = 0644,
1415                 .proc_handler   = proc_dointvec,
1416         },
1417 #endif
1418 #ifdef CONFIG_DNOTIFY
1419         {
1420                 .procname       = "dir-notify-enable",
1421                 .data           = &dir_notify_enable,
1422                 .maxlen         = sizeof(int),
1423                 .mode           = 0644,
1424                 .proc_handler   = proc_dointvec,
1425         },
1426 #endif
1427 #ifdef CONFIG_MMU
1428 #ifdef CONFIG_FILE_LOCKING
1429         {
1430                 .procname       = "lease-break-time",
1431                 .data           = &lease_break_time,
1432                 .maxlen         = sizeof(int),
1433                 .mode           = 0644,
1434                 .proc_handler   = proc_dointvec,
1435         },
1436 #endif
1437 #ifdef CONFIG_AIO
1438         {
1439                 .procname       = "aio-nr",
1440                 .data           = &aio_nr,
1441                 .maxlen         = sizeof(aio_nr),
1442                 .mode           = 0444,
1443                 .proc_handler   = proc_doulongvec_minmax,
1444         },
1445         {
1446                 .procname       = "aio-max-nr",
1447                 .data           = &aio_max_nr,
1448                 .maxlen         = sizeof(aio_max_nr),
1449                 .mode           = 0644,
1450                 .proc_handler   = proc_doulongvec_minmax,
1451         },
1452 #endif /* CONFIG_AIO */
1453 #ifdef CONFIG_INOTIFY_USER
1454         {
1455                 .procname       = "inotify",
1456                 .mode           = 0555,
1457                 .child          = inotify_table,
1458         },
1459 #endif
1460 #ifdef CONFIG_EPOLL
1461         {
1462                 .procname       = "epoll",
1463                 .mode           = 0555,
1464                 .child          = epoll_table,
1465         },
1466 #endif
1467 #endif
1468         {
1469                 .procname       = "suid_dumpable",
1470                 .data           = &suid_dumpable,
1471                 .maxlen         = sizeof(int),
1472                 .mode           = 0644,
1473                 .proc_handler   = proc_dointvec_minmax,
1474                 .extra1         = &zero,
1475                 .extra2         = &two,
1476         },
1477 #if defined(CONFIG_BINFMT_MISC) || defined(CONFIG_BINFMT_MISC_MODULE)
1478         {
1479                 .procname       = "binfmt_misc",
1480                 .mode           = 0555,
1481                 .child          = binfmt_misc_table,
1482         },
1483 #endif
1484         {
1485                 .procname       = "pipe-max-size",
1486                 .data           = &pipe_max_size,
1487                 .maxlen         = sizeof(int),
1488                 .mode           = 0644,
1489                 .proc_handler   = &pipe_proc_fn,
1490                 .extra1         = &pipe_min_size,
1491         },
1492 /*
1493  * NOTE: do not add new entries to this table unless you have read
1494  * Documentation/sysctl/ctl_unnumbered.txt
1495  */
1496         { }
1497 };
1498
1499 static struct ctl_table debug_table[] = {
1500 #if defined(CONFIG_X86) || defined(CONFIG_PPC) || defined(CONFIG_SPARC) || \
1501     defined(CONFIG_S390)
1502         {
1503                 .procname       = "exception-trace",
1504                 .data           = &show_unhandled_signals,
1505                 .maxlen         = sizeof(int),
1506                 .mode           = 0644,
1507                 .proc_handler   = proc_dointvec
1508         },
1509 #endif
1510 #if defined(CONFIG_OPTPROBES)
1511         {
1512                 .procname       = "kprobes-optimization",
1513                 .data           = &sysctl_kprobes_optimization,
1514                 .maxlen         = sizeof(int),
1515                 .mode           = 0644,
1516                 .proc_handler   = proc_kprobes_optimization_handler,
1517                 .extra1         = &zero,
1518                 .extra2         = &one,
1519         },
1520 #endif
1521         { }
1522 };
1523
1524 static struct ctl_table dev_table[] = {
1525         { }
1526 };
1527
1528 static DEFINE_SPINLOCK(sysctl_lock);
1529
1530 /* called under sysctl_lock */
1531 static int use_table(struct ctl_table_header *p)
1532 {
1533         if (unlikely(p->unregistering))
1534                 return 0;
1535         p->used++;
1536         return 1;
1537 }
1538
1539 /* called under sysctl_lock */
1540 static void unuse_table(struct ctl_table_header *p)
1541 {
1542         if (!--p->used)
1543                 if (unlikely(p->unregistering))
1544                         complete(p->unregistering);
1545 }
1546
1547 /* called under sysctl_lock, will reacquire if has to wait */
1548 static void start_unregistering(struct ctl_table_header *p)
1549 {
1550         /*
1551          * if p->used is 0, nobody will ever touch that entry again;
1552          * we'll eliminate all paths to it before dropping sysctl_lock
1553          */
1554         if (unlikely(p->used)) {
1555                 struct completion wait;
1556                 init_completion(&wait);
1557                 p->unregistering = &wait;
1558                 spin_unlock(&sysctl_lock);
1559                 wait_for_completion(&wait);
1560                 spin_lock(&sysctl_lock);
1561         } else {
1562                 /* anything non-NULL; we'll never dereference it */
1563                 p->unregistering = ERR_PTR(-EINVAL);
1564         }
1565         /*
1566          * do not remove from the list until nobody holds it; walking the
1567          * list in do_sysctl() relies on that.
1568          */
1569         list_del_init(&p->ctl_entry);
1570 }
1571
1572 void sysctl_head_get(struct ctl_table_header *head)
1573 {
1574         spin_lock(&sysctl_lock);
1575         head->count++;
1576         spin_unlock(&sysctl_lock);
1577 }
1578
1579 void sysctl_head_put(struct ctl_table_header *head)
1580 {
1581         spin_lock(&sysctl_lock);
1582         if (!--head->count)
1583                 kfree(head);
1584         spin_unlock(&sysctl_lock);
1585 }
1586
1587 struct ctl_table_header *sysctl_head_grab(struct ctl_table_header *head)
1588 {
1589         if (!head)
1590                 BUG();
1591         spin_lock(&sysctl_lock);
1592         if (!use_table(head))
1593                 head = ERR_PTR(-ENOENT);
1594         spin_unlock(&sysctl_lock);
1595         return head;
1596 }
1597
1598 void sysctl_head_finish(struct ctl_table_header *head)
1599 {
1600         if (!head)
1601                 return;
1602         spin_lock(&sysctl_lock);
1603         unuse_table(head);
1604         spin_unlock(&sysctl_lock);
1605 }
1606
1607 static struct ctl_table_set *
1608 lookup_header_set(struct ctl_table_root *root, struct nsproxy *namespaces)
1609 {
1610         struct ctl_table_set *set = &root->default_set;
1611         if (root->lookup)
1612                 set = root->lookup(root, namespaces);
1613         return set;
1614 }
1615
1616 static struct list_head *
1617 lookup_header_list(struct ctl_table_root *root, struct nsproxy *namespaces)
1618 {
1619         struct ctl_table_set *set = lookup_header_set(root, namespaces);
1620         return &set->list;
1621 }
1622
1623 struct ctl_table_header *__sysctl_head_next(struct nsproxy *namespaces,
1624                                             struct ctl_table_header *prev)
1625 {
1626         struct ctl_table_root *root;
1627         struct list_head *header_list;
1628         struct ctl_table_header *head;
1629         struct list_head *tmp;
1630
1631         spin_lock(&sysctl_lock);
1632         if (prev) {
1633                 head = prev;
1634                 tmp = &prev->ctl_entry;
1635                 unuse_table(prev);
1636                 goto next;
1637         }
1638         tmp = &root_table_header.ctl_entry;
1639         for (;;) {
1640                 head = list_entry(tmp, struct ctl_table_header, ctl_entry);
1641
1642                 if (!use_table(head))
1643                         goto next;
1644                 spin_unlock(&sysctl_lock);
1645                 return head;
1646         next:
1647                 root = head->root;
1648                 tmp = tmp->next;
1649                 header_list = lookup_header_list(root, namespaces);
1650                 if (tmp != header_list)
1651                         continue;
1652
1653                 do {
1654                         root = list_entry(root->root_list.next,
1655                                         struct ctl_table_root, root_list);
1656                         if (root == &sysctl_table_root)
1657                                 goto out;
1658                         header_list = lookup_header_list(root, namespaces);
1659                 } while (list_empty(header_list));
1660                 tmp = header_list->next;
1661         }
1662 out:
1663         spin_unlock(&sysctl_lock);
1664         return NULL;
1665 }
1666
1667 struct ctl_table_header *sysctl_head_next(struct ctl_table_header *prev)
1668 {
1669         return __sysctl_head_next(current->nsproxy, prev);
1670 }
1671
1672 void register_sysctl_root(struct ctl_table_root *root)
1673 {
1674         spin_lock(&sysctl_lock);
1675         list_add_tail(&root->root_list, &sysctl_table_root.root_list);
1676         spin_unlock(&sysctl_lock);
1677 }
1678
1679 /*
1680  * sysctl_perm does NOT grant the superuser all rights automatically, because
1681  * some sysctl variables are readonly even to root.
1682  */
1683
1684 static int test_perm(int mode, int op)
1685 {
1686         if (!current_euid())
1687                 mode >>= 6;
1688         else if (in_egroup_p(0))
1689                 mode >>= 3;
1690         if ((op & ~mode & (MAY_READ|MAY_WRITE|MAY_EXEC)) == 0)
1691                 return 0;
1692         return -EACCES;
1693 }
1694
1695 int sysctl_perm(struct ctl_table_root *root, struct ctl_table *table, int op)
1696 {
1697         int error;
1698         int mode;
1699
1700         error = security_sysctl(table, op & (MAY_READ | MAY_WRITE | MAY_EXEC));
1701         if (error)
1702                 return error;
1703
1704         if (root->permissions)
1705                 mode = root->permissions(root, current->nsproxy, table);
1706         else
1707                 mode = table->mode;
1708
1709         return test_perm(mode, op);
1710 }
1711
1712 static void sysctl_set_parent(struct ctl_table *parent, struct ctl_table *table)
1713 {
1714         for (; table->procname; table++) {
1715                 table->parent = parent;
1716                 if (table->child)
1717                         sysctl_set_parent(table, table->child);
1718         }
1719 }
1720
1721 static __init int sysctl_init(void)
1722 {
1723         sysctl_set_parent(NULL, root_table);
1724 #ifdef CONFIG_SYSCTL_SYSCALL_CHECK
1725         {
1726                 int err;
1727                 err = sysctl_check_table(current->nsproxy, root_table);
1728         }
1729 #endif
1730         return 0;
1731 }
1732
1733 core_initcall(sysctl_init);
1734
1735 static struct ctl_table *is_branch_in(struct ctl_table *branch,
1736                                       struct ctl_table *table)
1737 {
1738         struct ctl_table *p;
1739         const char *s = branch->procname;
1740
1741         /* branch should have named subdirectory as its first element */
1742         if (!s || !branch->child)
1743                 return NULL;
1744
1745         /* ... and nothing else */
1746         if (branch[1].procname)
1747                 return NULL;
1748
1749         /* table should contain subdirectory with the same name */
1750         for (p = table; p->procname; p++) {
1751                 if (!p->child)
1752                         continue;
1753                 if (p->procname && strcmp(p->procname, s) == 0)
1754                         return p;
1755         }
1756         return NULL;
1757 }
1758
1759 /* see if attaching q to p would be an improvement */
1760 static void try_attach(struct ctl_table_header *p, struct ctl_table_header *q)
1761 {
1762         struct ctl_table *to = p->ctl_table, *by = q->ctl_table;
1763         struct ctl_table *next;
1764         int is_better = 0;
1765         int not_in_parent = !p->attached_by;
1766
1767         while ((next = is_branch_in(by, to)) != NULL) {
1768                 if (by == q->attached_by)
1769                         is_better = 1;
1770                 if (to == p->attached_by)
1771                         not_in_parent = 1;
1772                 by = by->child;
1773                 to = next->child;
1774         }
1775
1776         if (is_better && not_in_parent) {
1777                 q->attached_by = by;
1778                 q->attached_to = to;
1779                 q->parent = p;
1780         }
1781 }
1782
1783 /**
1784  * __register_sysctl_paths - register a sysctl hierarchy
1785  * @root: List of sysctl headers to register on
1786  * @namespaces: Data to compute which lists of sysctl entries are visible
1787  * @path: The path to the directory the sysctl table is in.
1788  * @table: the top-level table structure
1789  *
1790  * Register a sysctl table hierarchy. @table should be a filled in ctl_table
1791  * array. A completely 0 filled entry terminates the table.
1792  *
1793  * The members of the &struct ctl_table structure are used as follows:
1794  *
1795  * procname - the name of the sysctl file under /proc/sys. Set to %NULL to not
1796  *            enter a sysctl file
1797  *
1798  * data - a pointer to data for use by proc_handler
1799  *
1800  * maxlen - the maximum size in bytes of the data
1801  *
1802  * mode - the file permissions for the /proc/sys file, and for sysctl(2)
1803  *
1804  * child - a pointer to the child sysctl table if this entry is a directory, or
1805  *         %NULL.
1806  *
1807  * proc_handler - the text handler routine (described below)
1808  *
1809  * de - for internal use by the sysctl routines
1810  *
1811  * extra1, extra2 - extra pointers usable by the proc handler routines
1812  *
1813  * Leaf nodes in the sysctl tree will be represented by a single file
1814  * under /proc; non-leaf nodes will be represented by directories.
1815  *
1816  * sysctl(2) can automatically manage read and write requests through
1817  * the sysctl table.  The data and maxlen fields of the ctl_table
1818  * struct enable minimal validation of the values being written to be
1819  * performed, and the mode field allows minimal authentication.
1820  *
1821  * There must be a proc_handler routine for any terminal nodes
1822  * mirrored under /proc/sys (non-terminals are handled by a built-in
1823  * directory handler).  Several default handlers are available to
1824  * cover common cases -
1825  *
1826  * proc_dostring(), proc_dointvec(), proc_dointvec_jiffies(),
1827  * proc_dointvec_userhz_jiffies(), proc_dointvec_minmax(),
1828  * proc_doulongvec_ms_jiffies_minmax(), proc_doulongvec_minmax()
1829  *
1830  * It is the handler's job to read the input buffer from user memory
1831  * and process it. The handler should return 0 on success.
1832  *
1833  * This routine returns %NULL on a failure to register, and a pointer
1834  * to the table header on success.
1835  */
1836 struct ctl_table_header *__register_sysctl_paths(
1837         struct ctl_table_root *root,
1838         struct nsproxy *namespaces,
1839         const struct ctl_path *path, struct ctl_table *table)
1840 {
1841         struct ctl_table_header *header;
1842         struct ctl_table *new, **prevp;
1843         unsigned int n, npath;
1844         struct ctl_table_set *set;
1845
1846         /* Count the path components */
1847         for (npath = 0; path[npath].procname; ++npath)
1848                 ;
1849
1850         /*
1851          * For each path component, allocate a 2-element ctl_table array.
1852          * The first array element will be filled with the sysctl entry
1853          * for this, the second will be the sentinel (procname == 0).
1854          *
1855          * We allocate everything in one go so that we don't have to
1856          * worry about freeing additional memory in unregister_sysctl_table.
1857          */
1858         header = kzalloc(sizeof(struct ctl_table_header) +
1859                          (2 * npath * sizeof(struct ctl_table)), GFP_KERNEL);
1860         if (!header)
1861                 return NULL;
1862
1863         new = (struct ctl_table *) (header + 1);
1864
1865         /* Now connect the dots */
1866         prevp = &header->ctl_table;
1867         for (n = 0; n < npath; ++n, ++path) {
1868                 /* Copy the procname */
1869                 new->procname = path->procname;
1870                 new->mode     = 0555;
1871
1872                 *prevp = new;
1873                 prevp = &new->child;
1874
1875                 new += 2;
1876         }
1877         *prevp = table;
1878         header->ctl_table_arg = table;
1879
1880         INIT_LIST_HEAD(&header->ctl_entry);
1881         header->used = 0;
1882         header->unregistering = NULL;
1883         header->root = root;
1884         sysctl_set_parent(NULL, header->ctl_table);
1885         header->count = 1;
1886 #ifdef CONFIG_SYSCTL_SYSCALL_CHECK
1887         if (sysctl_check_table(namespaces, header->ctl_table)) {
1888                 kfree(header);
1889                 return NULL;
1890         }
1891 #endif
1892         spin_lock(&sysctl_lock);
1893         header->set = lookup_header_set(root, namespaces);
1894         header->attached_by = header->ctl_table;
1895         header->attached_to = root_table;
1896         header->parent = &root_table_header;
1897         for (set = header->set; set; set = set->parent) {
1898                 struct ctl_table_header *p;
1899                 list_for_each_entry(p, &set->list, ctl_entry) {
1900                         if (p->unregistering)
1901                                 continue;
1902                         try_attach(p, header);
1903                 }
1904         }
1905         header->parent->count++;
1906         list_add_tail(&header->ctl_entry, &header->set->list);
1907         spin_unlock(&sysctl_lock);
1908
1909         return header;
1910 }
1911
1912 /**
1913  * register_sysctl_table_path - register a sysctl table hierarchy
1914  * @path: The path to the directory the sysctl table is in.
1915  * @table: the top-level table structure
1916  *
1917  * Register a sysctl table hierarchy. @table should be a filled in ctl_table
1918  * array. A completely 0 filled entry terminates the table.
1919  *
1920  * See __register_sysctl_paths for more details.
1921  */
1922 struct ctl_table_header *register_sysctl_paths(const struct ctl_path *path,
1923                                                 struct ctl_table *table)
1924 {
1925         return __register_sysctl_paths(&sysctl_table_root, current->nsproxy,
1926                                         path, table);
1927 }
1928
1929 /**
1930  * register_sysctl_table - register a sysctl table hierarchy
1931  * @table: the top-level table structure
1932  *
1933  * Register a sysctl table hierarchy. @table should be a filled in ctl_table
1934  * array. A completely 0 filled entry terminates the table.
1935  *
1936  * See register_sysctl_paths for more details.
1937  */
1938 struct ctl_table_header *register_sysctl_table(struct ctl_table *table)
1939 {
1940         static const struct ctl_path null_path[] = { {} };
1941
1942         return register_sysctl_paths(null_path, table);
1943 }
1944
1945 /**
1946  * unregister_sysctl_table - unregister a sysctl table hierarchy
1947  * @header: the header returned from register_sysctl_table
1948  *
1949  * Unregisters the sysctl table and all children. proc entries may not
1950  * actually be removed until they are no longer used by anyone.
1951  */
1952 void unregister_sysctl_table(struct ctl_table_header * header)
1953 {
1954         might_sleep();
1955
1956         if (header == NULL)
1957                 return;
1958
1959         spin_lock(&sysctl_lock);
1960         start_unregistering(header);
1961         if (!--header->parent->count) {
1962                 WARN_ON(1);
1963                 kfree(header->parent);
1964         }
1965         if (!--header->count)
1966                 kfree(header);
1967         spin_unlock(&sysctl_lock);
1968 }
1969
1970 int sysctl_is_seen(struct ctl_table_header *p)
1971 {
1972         struct ctl_table_set *set = p->set;
1973         int res;
1974         spin_lock(&sysctl_lock);
1975         if (p->unregistering)
1976                 res = 0;
1977         else if (!set->is_seen)
1978                 res = 1;
1979         else
1980                 res = set->is_seen(set);
1981         spin_unlock(&sysctl_lock);
1982         return res;
1983 }
1984
1985 void setup_sysctl_set(struct ctl_table_set *p,
1986         struct ctl_table_set *parent,
1987         int (*is_seen)(struct ctl_table_set *))
1988 {
1989         INIT_LIST_HEAD(&p->list);
1990         p->parent = parent ? parent : &sysctl_table_root.default_set;
1991         p->is_seen = is_seen;
1992 }
1993
1994 #else /* !CONFIG_SYSCTL */
1995 struct ctl_table_header *register_sysctl_table(struct ctl_table * table)
1996 {
1997         return NULL;
1998 }
1999
2000 struct ctl_table_header *register_sysctl_paths(const struct ctl_path *path,
2001                                                     struct ctl_table *table)
2002 {
2003         return NULL;
2004 }
2005
2006 void unregister_sysctl_table(struct ctl_table_header * table)
2007 {
2008 }
2009
2010 void setup_sysctl_set(struct ctl_table_set *p,
2011         struct ctl_table_set *parent,
2012         int (*is_seen)(struct ctl_table_set *))
2013 {
2014 }
2015
2016 void sysctl_head_put(struct ctl_table_header *head)
2017 {
2018 }
2019
2020 #endif /* CONFIG_SYSCTL */
2021
2022 /*
2023  * /proc/sys support
2024  */
2025
2026 #ifdef CONFIG_PROC_SYSCTL
2027
2028 static int _proc_do_string(void* data, int maxlen, int write,
2029                            void __user *buffer,
2030                            size_t *lenp, loff_t *ppos)
2031 {
2032         size_t len;
2033         char __user *p;
2034         char c;
2035
2036         if (!data || !maxlen || !*lenp) {
2037                 *lenp = 0;
2038                 return 0;
2039         }
2040
2041         if (write) {
2042                 len = 0;
2043                 p = buffer;
2044                 while (len < *lenp) {
2045                         if (get_user(c, p++))
2046                                 return -EFAULT;
2047                         if (c == 0 || c == '\n')
2048                                 break;
2049                         len++;
2050                 }
2051                 if (len >= maxlen)
2052                         len = maxlen-1;
2053                 if(copy_from_user(data, buffer, len))
2054                         return -EFAULT;
2055                 ((char *) data)[len] = 0;
2056                 *ppos += *lenp;
2057         } else {
2058                 len = strlen(data);
2059                 if (len > maxlen)
2060                         len = maxlen;
2061
2062                 if (*ppos > len) {
2063                         *lenp = 0;
2064                         return 0;
2065                 }
2066
2067                 data += *ppos;
2068                 len  -= *ppos;
2069
2070                 if (len > *lenp)
2071                         len = *lenp;
2072                 if (len)
2073                         if(copy_to_user(buffer, data, len))
2074                                 return -EFAULT;
2075                 if (len < *lenp) {
2076                         if(put_user('\n', ((char __user *) buffer) + len))
2077                                 return -EFAULT;
2078                         len++;
2079                 }
2080                 *lenp = len;
2081                 *ppos += len;
2082         }
2083         return 0;
2084 }
2085
2086 /**
2087  * proc_dostring - read a string sysctl
2088  * @table: the sysctl table
2089  * @write: %TRUE if this is a write to the sysctl file
2090  * @buffer: the user buffer
2091  * @lenp: the size of the user buffer
2092  * @ppos: file position
2093  *
2094  * Reads/writes a string from/to the user buffer. If the kernel
2095  * buffer provided is not large enough to hold the string, the
2096  * string is truncated. The copied string is %NULL-terminated.
2097  * If the string is being read by the user process, it is copied
2098  * and a newline '\n' is added. It is truncated if the buffer is
2099  * not large enough.
2100  *
2101  * Returns 0 on success.
2102  */
2103 int proc_dostring(struct ctl_table *table, int write,
2104                   void __user *buffer, size_t *lenp, loff_t *ppos)
2105 {
2106         return _proc_do_string(table->data, table->maxlen, write,
2107                                buffer, lenp, ppos);
2108 }
2109
2110 static size_t proc_skip_spaces(char **buf)
2111 {
2112         size_t ret;
2113         char *tmp = skip_spaces(*buf);
2114         ret = tmp - *buf;
2115         *buf = tmp;
2116         return ret;
2117 }
2118
2119 static void proc_skip_char(char **buf, size_t *size, const char v)
2120 {
2121         while (*size) {
2122                 if (**buf != v)
2123                         break;
2124                 (*size)--;
2125                 (*buf)++;
2126         }
2127 }
2128
2129 #define TMPBUFLEN 22
2130 /**
2131  * proc_get_long - reads an ASCII formatted integer from a user buffer
2132  *
2133  * @buf: a kernel buffer
2134  * @size: size of the kernel buffer
2135  * @val: this is where the number will be stored
2136  * @neg: set to %TRUE if number is negative
2137  * @perm_tr: a vector which contains the allowed trailers
2138  * @perm_tr_len: size of the perm_tr vector
2139  * @tr: pointer to store the trailer character
2140  *
2141  * In case of success %0 is returned and @buf and @size are updated with
2142  * the amount of bytes read. If @tr is non-NULL and a trailing
2143  * character exists (size is non-zero after returning from this
2144  * function), @tr is updated with the trailing character.
2145  */
2146 static int proc_get_long(char **buf, size_t *size,
2147                           unsigned long *val, bool *neg,
2148                           const char *perm_tr, unsigned perm_tr_len, char *tr)
2149 {
2150         int len;
2151         char *p, tmp[TMPBUFLEN];
2152
2153         if (!*size)
2154                 return -EINVAL;
2155
2156         len = *size;
2157         if (len > TMPBUFLEN - 1)
2158                 len = TMPBUFLEN - 1;
2159
2160         memcpy(tmp, *buf, len);
2161
2162         tmp[len] = 0;
2163         p = tmp;
2164         if (*p == '-' && *size > 1) {
2165                 *neg = true;
2166                 p++;
2167         } else
2168                 *neg = false;
2169         if (!isdigit(*p))
2170                 return -EINVAL;
2171
2172         *val = simple_strtoul(p, &p, 0);
2173
2174         len = p - tmp;
2175
2176         /* We don't know if the next char is whitespace thus we may accept
2177          * invalid integers (e.g. 1234...a) or two integers instead of one
2178          * (e.g. 123...1). So lets not allow such large numbers. */
2179         if (len == TMPBUFLEN - 1)
2180                 return -EINVAL;
2181
2182         if (len < *size && perm_tr_len && !memchr(perm_tr, *p, perm_tr_len))
2183                 return -EINVAL;
2184
2185         if (tr && (len < *size))
2186                 *tr = *p;
2187
2188         *buf += len;
2189         *size -= len;
2190
2191         return 0;
2192 }
2193
2194 /**
2195  * proc_put_long - converts an integer to a decimal ASCII formatted string
2196  *
2197  * @buf: the user buffer
2198  * @size: the size of the user buffer
2199  * @val: the integer to be converted
2200  * @neg: sign of the number, %TRUE for negative
2201  *
2202  * In case of success %0 is returned and @buf and @size are updated with
2203  * the amount of bytes written.
2204  */
2205 static int proc_put_long(void __user **buf, size_t *size, unsigned long val,
2206                           bool neg)
2207 {
2208         int len;
2209         char tmp[TMPBUFLEN], *p = tmp;
2210
2211         sprintf(p, "%s%lu", neg ? "-" : "", val);
2212         len = strlen(tmp);
2213         if (len > *size)
2214                 len = *size;
2215         if (copy_to_user(*buf, tmp, len))
2216                 return -EFAULT;
2217         *size -= len;
2218         *buf += len;
2219         return 0;
2220 }
2221 #undef TMPBUFLEN
2222
2223 static int proc_put_char(void __user **buf, size_t *size, char c)
2224 {
2225         if (*size) {
2226                 char __user **buffer = (char __user **)buf;
2227                 if (put_user(c, *buffer))
2228                         return -EFAULT;
2229                 (*size)--, (*buffer)++;
2230                 *buf = *buffer;
2231         }
2232         return 0;
2233 }
2234
2235 static int do_proc_dointvec_conv(bool *negp, unsigned long *lvalp,
2236                                  int *valp,
2237                                  int write, void *data)
2238 {
2239         if (write) {
2240                 *valp = *negp ? -*lvalp : *lvalp;
2241         } else {
2242                 int val = *valp;
2243                 if (val < 0) {
2244                         *negp = true;
2245                         *lvalp = (unsigned long)-val;
2246                 } else {
2247                         *negp = false;
2248                         *lvalp = (unsigned long)val;
2249                 }
2250         }
2251         return 0;
2252 }
2253
2254 static const char proc_wspace_sep[] = { ' ', '\t', '\n' };
2255
2256 static int __do_proc_dointvec(void *tbl_data, struct ctl_table *table,
2257                   int write, void __user *buffer,
2258                   size_t *lenp, loff_t *ppos,
2259                   int (*conv)(bool *negp, unsigned long *lvalp, int *valp,
2260                               int write, void *data),
2261                   void *data)
2262 {
2263         int *i, vleft, first = 1, err = 0;
2264         unsigned long page = 0;
2265         size_t left;
2266         char *kbuf;
2267
2268         if (!tbl_data || !table->maxlen || !*lenp || (*ppos && !write)) {
2269                 *lenp = 0;
2270                 return 0;
2271         }
2272
2273         i = (int *) tbl_data;
2274         vleft = table->maxlen / sizeof(*i);
2275         left = *lenp;
2276
2277         if (!conv)
2278                 conv = do_proc_dointvec_conv;
2279
2280         if (write) {
2281                 if (left > PAGE_SIZE - 1)
2282                         left = PAGE_SIZE - 1;
2283                 page = __get_free_page(GFP_TEMPORARY);
2284                 kbuf = (char *) page;
2285                 if (!kbuf)
2286                         return -ENOMEM;
2287                 if (copy_from_user(kbuf, buffer, left)) {
2288                         err = -EFAULT;
2289                         goto free;
2290                 }
2291                 kbuf[left] = 0;
2292         }
2293
2294         for (; left && vleft--; i++, first=0) {
2295                 unsigned long lval;
2296                 bool neg;
2297
2298                 if (write) {
2299                         left -= proc_skip_spaces(&kbuf);
2300
2301                         if (!left)
2302                                 break;
2303                         err = proc_get_long(&kbuf, &left, &lval, &neg,
2304                                              proc_wspace_sep,
2305                                              sizeof(proc_wspace_sep), NULL);
2306                         if (err)
2307                                 break;
2308                         if (conv(&neg, &lval, i, 1, data)) {
2309                                 err = -EINVAL;
2310                                 break;
2311                         }
2312                 } else {
2313                         if (conv(&neg, &lval, i, 0, data)) {
2314                                 err = -EINVAL;
2315                                 break;
2316                         }
2317                         if (!first)
2318                                 err = proc_put_char(&buffer, &left, '\t');
2319                         if (err)
2320                                 break;
2321                         err = proc_put_long(&buffer, &left, lval, neg);
2322                         if (err)
2323                                 break;
2324                 }
2325         }
2326
2327         if (!write && !first && left && !err)
2328                 err = proc_put_char(&buffer, &left, '\n');
2329         if (write && !err && left)
2330                 left -= proc_skip_spaces(&kbuf);
2331 free:
2332         if (write) {
2333                 free_page(page);
2334                 if (first)
2335                         return err ? : -EINVAL;
2336         }
2337         *lenp -= left;
2338         *ppos += *lenp;
2339         return err;
2340 }
2341
2342 static int do_proc_dointvec(struct ctl_table *table, int write,
2343                   void __user *buffer, size_t *lenp, loff_t *ppos,
2344                   int (*conv)(bool *negp, unsigned long *lvalp, int *valp,
2345                               int write, void *data),
2346                   void *data)
2347 {
2348         return __do_proc_dointvec(table->data, table, write,
2349                         buffer, lenp, ppos, conv, data);
2350 }
2351
2352 /**
2353  * proc_dointvec - read a vector of integers
2354  * @table: the sysctl table
2355  * @write: %TRUE if this is a write to the sysctl file
2356  * @buffer: the user buffer
2357  * @lenp: the size of the user buffer
2358  * @ppos: file position
2359  *
2360  * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2361  * values from/to the user buffer, treated as an ASCII string.
2362  *
2363  * Returns 0 on success.
2364  */
2365 int proc_dointvec(struct ctl_table *table, int write,
2366                      void __user *buffer, size_t *lenp, loff_t *ppos)
2367 {
2368     return do_proc_dointvec(table,write,buffer,lenp,ppos,
2369                             NULL,NULL);
2370 }
2371
2372 /*
2373  * Taint values can only be increased
2374  * This means we can safely use a temporary.
2375  */
2376 static int proc_taint(struct ctl_table *table, int write,
2377                                void __user *buffer, size_t *lenp, loff_t *ppos)
2378 {
2379         struct ctl_table t;
2380         unsigned long tmptaint = get_taint();
2381         int err;
2382
2383         if (write && !capable(CAP_SYS_ADMIN))
2384                 return -EPERM;
2385
2386         t = *table;
2387         t.data = &tmptaint;
2388         err = proc_doulongvec_minmax(&t, write, buffer, lenp, ppos);
2389         if (err < 0)
2390                 return err;
2391
2392         if (write) {
2393                 /*
2394                  * Poor man's atomic or. Not worth adding a primitive
2395                  * to everyone's atomic.h for this
2396                  */
2397                 int i;
2398                 for (i = 0; i < BITS_PER_LONG && tmptaint >> i; i++) {
2399                         if ((tmptaint >> i) & 1)
2400                                 add_taint(i);
2401                 }
2402         }
2403
2404         return err;
2405 }
2406
2407 struct do_proc_dointvec_minmax_conv_param {
2408         int *min;
2409         int *max;
2410 };
2411
2412 static int do_proc_dointvec_minmax_conv(bool *negp, unsigned long *lvalp,
2413                                         int *valp,
2414                                         int write, void *data)
2415 {
2416         struct do_proc_dointvec_minmax_conv_param *param = data;
2417         if (write) {
2418                 int val = *negp ? -*lvalp : *lvalp;
2419                 if ((param->min && *param->min > val) ||
2420                     (param->max && *param->max < val))
2421                         return -EINVAL;
2422                 *valp = val;
2423         } else {
2424                 int val = *valp;
2425                 if (val < 0) {
2426                         *negp = true;
2427                         *lvalp = (unsigned long)-val;
2428                 } else {
2429                         *negp = false;
2430                         *lvalp = (unsigned long)val;
2431                 }
2432         }
2433         return 0;
2434 }
2435
2436 /**
2437  * proc_dointvec_minmax - read a vector of integers with min/max values
2438  * @table: the sysctl table
2439  * @write: %TRUE if this is a write to the sysctl file
2440  * @buffer: the user buffer
2441  * @lenp: the size of the user buffer
2442  * @ppos: file position
2443  *
2444  * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2445  * values from/to the user buffer, treated as an ASCII string.
2446  *
2447  * This routine will ensure the values are within the range specified by
2448  * table->extra1 (min) and table->extra2 (max).
2449  *
2450  * Returns 0 on success.
2451  */
2452 int proc_dointvec_minmax(struct ctl_table *table, int write,
2453                   void __user *buffer, size_t *lenp, loff_t *ppos)
2454 {
2455         struct do_proc_dointvec_minmax_conv_param param = {
2456                 .min = (int *) table->extra1,
2457                 .max = (int *) table->extra2,
2458         };
2459         return do_proc_dointvec(table, write, buffer, lenp, ppos,
2460                                 do_proc_dointvec_minmax_conv, &param);
2461 }
2462
2463 static int __do_proc_doulongvec_minmax(void *data, struct ctl_table *table, int write,
2464                                      void __user *buffer,
2465                                      size_t *lenp, loff_t *ppos,
2466                                      unsigned long convmul,
2467                                      unsigned long convdiv)
2468 {
2469         unsigned long *i, *min, *max;
2470         int vleft, first = 1, err = 0;
2471         unsigned long page = 0;
2472         size_t left;
2473         char *kbuf;
2474
2475         if (!data || !table->maxlen || !*lenp || (*ppos && !write)) {
2476                 *lenp = 0;
2477                 return 0;
2478         }
2479
2480         i = (unsigned long *) data;
2481         min = (unsigned long *) table->extra1;
2482         max = (unsigned long *) table->extra2;
2483         vleft = table->maxlen / sizeof(unsigned long);
2484         left = *lenp;
2485
2486         if (write) {
2487                 if (left > PAGE_SIZE - 1)
2488                         left = PAGE_SIZE - 1;
2489                 page = __get_free_page(GFP_TEMPORARY);
2490                 kbuf = (char *) page;
2491                 if (!kbuf)
2492                         return -ENOMEM;
2493                 if (copy_from_user(kbuf, buffer, left)) {
2494                         err = -EFAULT;
2495                         goto free;
2496                 }
2497                 kbuf[left] = 0;
2498         }
2499
2500         for (; left && vleft--; i++, min++, max++, first=0) {
2501                 unsigned long val;
2502
2503                 if (write) {
2504                         bool neg;
2505
2506                         left -= proc_skip_spaces(&kbuf);
2507
2508                         err = proc_get_long(&kbuf, &left, &val, &neg,
2509                                              proc_wspace_sep,
2510                                              sizeof(proc_wspace_sep), NULL);
2511                         if (err)
2512                                 break;
2513                         if (neg)
2514                                 continue;
2515                         if ((min && val < *min) || (max && val > *max))
2516                                 continue;
2517                         *i = val;
2518                 } else {
2519                         val = convdiv * (*i) / convmul;
2520                         if (!first)
2521                                 err = proc_put_char(&buffer, &left, '\t');
2522                         err = proc_put_long(&buffer, &left, val, false);
2523                         if (err)
2524                                 break;
2525                 }
2526         }
2527
2528         if (!write && !first && left && !err)
2529                 err = proc_put_char(&buffer, &left, '\n');
2530         if (write && !err)
2531                 left -= proc_skip_spaces(&kbuf);
2532 free:
2533         if (write) {
2534                 free_page(page);
2535                 if (first)
2536                         return err ? : -EINVAL;
2537         }
2538         *lenp -= left;
2539         *ppos += *lenp;
2540         return err;
2541 }
2542
2543 static int do_proc_doulongvec_minmax(struct ctl_table *table, int write,
2544                                      void __user *buffer,
2545                                      size_t *lenp, loff_t *ppos,
2546                                      unsigned long convmul,
2547                                      unsigned long convdiv)
2548 {
2549         return __do_proc_doulongvec_minmax(table->data, table, write,
2550                         buffer, lenp, ppos, convmul, convdiv);
2551 }
2552
2553 /**
2554  * proc_doulongvec_minmax - read a vector of long integers with min/max values
2555  * @table: the sysctl table
2556  * @write: %TRUE if this is a write to the sysctl file
2557  * @buffer: the user buffer
2558  * @lenp: the size of the user buffer
2559  * @ppos: file position
2560  *
2561  * Reads/writes up to table->maxlen/sizeof(unsigned long) unsigned long
2562  * values from/to the user buffer, treated as an ASCII string.
2563  *
2564  * This routine will ensure the values are within the range specified by
2565  * table->extra1 (min) and table->extra2 (max).
2566  *
2567  * Returns 0 on success.
2568  */
2569 int proc_doulongvec_minmax(struct ctl_table *table, int write,
2570                            void __user *buffer, size_t *lenp, loff_t *ppos)
2571 {
2572     return do_proc_doulongvec_minmax(table, write, buffer, lenp, ppos, 1l, 1l);
2573 }
2574
2575 /**
2576  * proc_doulongvec_ms_jiffies_minmax - read a vector of millisecond values with min/max values
2577  * @table: the sysctl table
2578  * @write: %TRUE if this is a write to the sysctl file
2579  * @buffer: the user buffer
2580  * @lenp: the size of the user buffer
2581  * @ppos: file position
2582  *
2583  * Reads/writes up to table->maxlen/sizeof(unsigned long) unsigned long
2584  * values from/to the user buffer, treated as an ASCII string. The values
2585  * are treated as milliseconds, and converted to jiffies when they are stored.
2586  *
2587  * This routine will ensure the values are within the range specified by
2588  * table->extra1 (min) and table->extra2 (max).
2589  *
2590  * Returns 0 on success.
2591  */
2592 int proc_doulongvec_ms_jiffies_minmax(struct ctl_table *table, int write,
2593                                       void __user *buffer,
2594                                       size_t *lenp, loff_t *ppos)
2595 {
2596     return do_proc_doulongvec_minmax(table, write, buffer,
2597                                      lenp, ppos, HZ, 1000l);
2598 }
2599
2600
2601 static int do_proc_dointvec_jiffies_conv(bool *negp, unsigned long *lvalp,
2602                                          int *valp,
2603                                          int write, void *data)
2604 {
2605         if (write) {
2606                 if (*lvalp > LONG_MAX / HZ)
2607                         return 1;
2608                 *valp = *negp ? -(*lvalp*HZ) : (*lvalp*HZ);
2609         } else {
2610                 int val = *valp;
2611                 unsigned long lval;
2612                 if (val < 0) {
2613                         *negp = true;
2614                         lval = (unsigned long)-val;
2615                 } else {
2616                         *negp = false;
2617                         lval = (unsigned long)val;
2618                 }
2619                 *lvalp = lval / HZ;
2620         }
2621         return 0;
2622 }
2623
2624 static int do_proc_dointvec_userhz_jiffies_conv(bool *negp, unsigned long *lvalp,
2625                                                 int *valp,
2626                                                 int write, void *data)
2627 {
2628         if (write) {
2629                 if (USER_HZ < HZ && *lvalp > (LONG_MAX / HZ) * USER_HZ)
2630                         return 1;
2631                 *valp = clock_t_to_jiffies(*negp ? -*lvalp : *lvalp);
2632         } else {
2633                 int val = *valp;
2634                 unsigned long lval;
2635                 if (val < 0) {
2636                         *negp = true;
2637                         lval = (unsigned long)-val;
2638                 } else {
2639                         *negp = false;
2640                         lval = (unsigned long)val;
2641                 }
2642                 *lvalp = jiffies_to_clock_t(lval);
2643         }
2644         return 0;
2645 }
2646
2647 static int do_proc_dointvec_ms_jiffies_conv(bool *negp, unsigned long *lvalp,
2648                                             int *valp,
2649                                             int write, void *data)
2650 {
2651         if (write) {
2652                 *valp = msecs_to_jiffies(*negp ? -*lvalp : *lvalp);
2653         } else {
2654                 int val = *valp;
2655                 unsigned long lval;
2656                 if (val < 0) {
2657                         *negp = true;
2658                         lval = (unsigned long)-val;
2659                 } else {
2660                         *negp = false;
2661                         lval = (unsigned long)val;
2662                 }
2663                 *lvalp = jiffies_to_msecs(lval);
2664         }
2665         return 0;
2666 }
2667
2668 /**
2669  * proc_dointvec_jiffies - read a vector of integers as seconds
2670  * @table: the sysctl table
2671  * @write: %TRUE if this is a write to the sysctl file
2672  * @buffer: the user buffer
2673  * @lenp: the size of the user buffer
2674  * @ppos: file position
2675  *
2676  * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2677  * values from/to the user buffer, treated as an ASCII string.
2678  * The values read are assumed to be in seconds, and are converted into
2679  * jiffies.
2680  *
2681  * Returns 0 on success.
2682  */
2683 int proc_dointvec_jiffies(struct ctl_table *table, int write,
2684                           void __user *buffer, size_t *lenp, loff_t *ppos)
2685 {
2686     return do_proc_dointvec(table,write,buffer,lenp,ppos,
2687                             do_proc_dointvec_jiffies_conv,NULL);
2688 }
2689
2690 /**
2691  * proc_dointvec_userhz_jiffies - read a vector of integers as 1/USER_HZ seconds
2692  * @table: the sysctl table
2693  * @write: %TRUE if this is a write to the sysctl file
2694  * @buffer: the user buffer
2695  * @lenp: the size of the user buffer
2696  * @ppos: pointer to the file position
2697  *
2698  * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2699  * values from/to the user buffer, treated as an ASCII string.
2700  * The values read are assumed to be in 1/USER_HZ seconds, and
2701  * are converted into jiffies.
2702  *
2703  * Returns 0 on success.
2704  */
2705 int proc_dointvec_userhz_jiffies(struct ctl_table *table, int write,
2706                                  void __user *buffer, size_t *lenp, loff_t *ppos)
2707 {
2708     return do_proc_dointvec(table,write,buffer,lenp,ppos,
2709                             do_proc_dointvec_userhz_jiffies_conv,NULL);
2710 }
2711
2712 /**
2713  * proc_dointvec_ms_jiffies - read a vector of integers as 1 milliseconds
2714  * @table: the sysctl table
2715  * @write: %TRUE if this is a write to the sysctl file
2716  * @buffer: the user buffer
2717  * @lenp: the size of the user buffer
2718  * @ppos: file position
2719  * @ppos: the current position in the file
2720  *
2721  * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2722  * values from/to the user buffer, treated as an ASCII string.
2723  * The values read are assumed to be in 1/1000 seconds, and
2724  * are converted into jiffies.
2725  *
2726  * Returns 0 on success.
2727  */
2728 int proc_dointvec_ms_jiffies(struct ctl_table *table, int write,
2729                              void __user *buffer, size_t *lenp, loff_t *ppos)
2730 {
2731         return do_proc_dointvec(table, write, buffer, lenp, ppos,
2732                                 do_proc_dointvec_ms_jiffies_conv, NULL);
2733 }
2734
2735 static int proc_do_cad_pid(struct ctl_table *table, int write,
2736                            void __user *buffer, size_t *lenp, loff_t *ppos)
2737 {
2738         struct pid *new_pid;
2739         pid_t tmp;
2740         int r;
2741
2742         tmp = pid_vnr(cad_pid);
2743
2744         r = __do_proc_dointvec(&tmp, table, write, buffer,
2745                                lenp, ppos, NULL, NULL);
2746         if (r || !write)
2747                 return r;
2748
2749         new_pid = find_get_pid(tmp);
2750         if (!new_pid)
2751                 return -ESRCH;
2752
2753         put_pid(xchg(&cad_pid, new_pid));
2754         return 0;
2755 }
2756
2757 /**
2758  * proc_do_large_bitmap - read/write from/to a large bitmap
2759  * @table: the sysctl table
2760  * @write: %TRUE if this is a write to the sysctl file
2761  * @buffer: the user buffer
2762  * @lenp: the size of the user buffer
2763  * @ppos: file position
2764  *
2765  * The bitmap is stored at table->data and the bitmap length (in bits)
2766  * in table->maxlen.
2767  *
2768  * We use a range comma separated format (e.g. 1,3-4,10-10) so that
2769  * large bitmaps may be represented in a compact manner. Writing into
2770  * the file will clear the bitmap then update it with the given input.
2771  *
2772  * Returns 0 on success.
2773  */
2774 int proc_do_large_bitmap(struct ctl_table *table, int write,
2775                          void __user *buffer, size_t *lenp, loff_t *ppos)
2776 {
2777         int err = 0;
2778         bool first = 1;
2779         size_t left = *lenp;
2780         unsigned long bitmap_len = table->maxlen;
2781         unsigned long *bitmap = (unsigned long *) table->data;
2782         unsigned long *tmp_bitmap = NULL;
2783         char tr_a[] = { '-', ',', '\n' }, tr_b[] = { ',', '\n', 0 }, c;
2784
2785         if (!bitmap_len || !left || (*ppos && !write)) {
2786                 *lenp = 0;
2787                 return 0;
2788         }
2789
2790         if (write) {
2791                 unsigned long page = 0;
2792                 char *kbuf;
2793
2794                 if (left > PAGE_SIZE - 1)
2795                         left = PAGE_SIZE - 1;
2796
2797                 page = __get_free_page(GFP_TEMPORARY);
2798                 kbuf = (char *) page;
2799                 if (!kbuf)
2800                         return -ENOMEM;
2801                 if (copy_from_user(kbuf, buffer, left)) {
2802                         free_page(page);
2803                         return -EFAULT;
2804                 }
2805                 kbuf[left] = 0;
2806
2807                 tmp_bitmap = kzalloc(BITS_TO_LONGS(bitmap_len) * sizeof(unsigned long),
2808                                      GFP_KERNEL);
2809                 if (!tmp_bitmap) {
2810                         free_page(page);
2811                         return -ENOMEM;
2812                 }
2813                 proc_skip_char(&kbuf, &left, '\n');
2814                 while (!err && left) {
2815                         unsigned long val_a, val_b;
2816                         bool neg;
2817
2818                         err = proc_get_long(&kbuf, &left, &val_a, &neg, tr_a,
2819                                              sizeof(tr_a), &c);
2820                         if (err)
2821                                 break;
2822                         if (val_a >= bitmap_len || neg) {
2823                                 err = -EINVAL;
2824                                 break;
2825                         }
2826
2827                         val_b = val_a;
2828                         if (left) {
2829                                 kbuf++;
2830                                 left--;
2831                         }
2832
2833                         if (c == '-') {
2834                                 err = proc_get_long(&kbuf, &left, &val_b,
2835                                                      &neg, tr_b, sizeof(tr_b),
2836                                                      &c);
2837                                 if (err)
2838                                         break;
2839                                 if (val_b >= bitmap_len || neg ||
2840                                     val_a > val_b) {
2841                                         err = -EINVAL;
2842                                         break;
2843                                 }
2844                                 if (left) {
2845                                         kbuf++;
2846                                         left--;
2847                                 }
2848                         }
2849
2850                         while (val_a <= val_b)
2851                                 set_bit(val_a++, tmp_bitmap);
2852
2853                         first = 0;
2854                         proc_skip_char(&kbuf, &left, '\n');
2855                 }
2856                 free_page(page);
2857         } else {
2858                 unsigned long bit_a, bit_b = 0;
2859
2860                 while (left) {
2861                         bit_a = find_next_bit(bitmap, bitmap_len, bit_b);
2862                         if (bit_a >= bitmap_len)
2863                                 break;
2864                         bit_b = find_next_zero_bit(bitmap, bitmap_len,
2865                                                    bit_a + 1) - 1;
2866
2867                         if (!first) {
2868                                 err = proc_put_char(&buffer, &left, ',');
2869                                 if (err)
2870                                         break;
2871                         }
2872                         err = proc_put_long(&buffer, &left, bit_a, false);
2873                         if (err)
2874                                 break;
2875                         if (bit_a != bit_b) {
2876                                 err = proc_put_char(&buffer, &left, '-');
2877                                 if (err)
2878                                         break;
2879                                 err = proc_put_long(&buffer, &left, bit_b, false);
2880                                 if (err)
2881                                         break;
2882                         }
2883
2884                         first = 0; bit_b++;
2885                 }
2886                 if (!err)
2887                         err = proc_put_char(&buffer, &left, '\n');
2888         }
2889
2890         if (!err) {
2891                 if (write) {
2892                         if (*ppos)
2893                                 bitmap_or(bitmap, bitmap, tmp_bitmap, bitmap_len);
2894                         else
2895                                 memcpy(bitmap, tmp_bitmap,
2896                                         BITS_TO_LONGS(bitmap_len) * sizeof(unsigned long));
2897                 }
2898                 kfree(tmp_bitmap);
2899                 *lenp -= left;
2900                 *ppos += *lenp;
2901                 return 0;
2902         } else {
2903                 kfree(tmp_bitmap);
2904                 return err;
2905         }
2906 }
2907
2908 #else /* CONFIG_PROC_FS */
2909
2910 int proc_dostring(struct ctl_table *table, int write,
2911                   void __user *buffer, size_t *lenp, loff_t *ppos)
2912 {
2913         return -ENOSYS;
2914 }
2915
2916 int proc_dointvec(struct ctl_table *table, int write,
2917                   void __user *buffer, size_t *lenp, loff_t *ppos)
2918 {
2919         return -ENOSYS;
2920 }
2921
2922 int proc_dointvec_minmax(struct ctl_table *table, int write,
2923                     void __user *buffer, size_t *lenp, loff_t *ppos)
2924 {
2925         return -ENOSYS;
2926 }
2927
2928 int proc_dointvec_jiffies(struct ctl_table *table, int write,
2929                     void __user *buffer, size_t *lenp, loff_t *ppos)
2930 {
2931         return -ENOSYS;
2932 }
2933
2934 int proc_dointvec_userhz_jiffies(struct ctl_table *table, int write,
2935                     void __user *buffer, size_t *lenp, loff_t *ppos)
2936 {
2937         return -ENOSYS;
2938 }
2939
2940 int proc_dointvec_ms_jiffies(struct ctl_table *table, int write,
2941                              void __user *buffer, size_t *lenp, loff_t *ppos)
2942 {
2943         return -ENOSYS;
2944 }
2945
2946 int proc_doulongvec_minmax(struct ctl_table *table, int write,
2947                     void __user *buffer, size_t *lenp, loff_t *ppos)
2948 {
2949         return -ENOSYS;
2950 }
2951
2952 int proc_doulongvec_ms_jiffies_minmax(struct ctl_table *table, int write,
2953                                       void __user *buffer,
2954                                       size_t *lenp, loff_t *ppos)
2955 {
2956     return -ENOSYS;
2957 }
2958
2959
2960 #endif /* CONFIG_PROC_FS */
2961
2962 /*
2963  * No sense putting this after each symbol definition, twice,
2964  * exception granted :-)
2965  */
2966 EXPORT_SYMBOL(proc_dointvec);
2967 EXPORT_SYMBOL(proc_dointvec_jiffies);
2968 EXPORT_SYMBOL(proc_dointvec_minmax);
2969 EXPORT_SYMBOL(proc_dointvec_userhz_jiffies);
2970 EXPORT_SYMBOL(proc_dointvec_ms_jiffies);
2971 EXPORT_SYMBOL(proc_dostring);
2972 EXPORT_SYMBOL(proc_doulongvec_minmax);
2973 EXPORT_SYMBOL(proc_doulongvec_ms_jiffies_minmax);
2974 EXPORT_SYMBOL(register_sysctl_table);
2975 EXPORT_SYMBOL(register_sysctl_paths);
2976 EXPORT_SYMBOL(unregister_sysctl_table);