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/printk.h>
  28 #include <linux/proc_fs.h>
  29 #include <linux/security.h>
  30 #include <linux/ctype.h>
  31 #include <linux/kmemcheck.h>
  32 #include <linux/fs.h>
  33 #include <linux/init.h>
  34 #include <linux/kernel.h>
  35 #include <linux/kobject.h>
  36 #include <linux/net.h>
  37 #include <linux/sysrq.h>
  38 #include <linux/highuid.h>
  39 #include <linux/writeback.h>
  40 #include <linux/ratelimit.h>
  41 #include <linux/compaction.h>
  42 #include <linux/hugetlb.h>
  43 #include <linux/initrd.h>
  44 #include <linux/key.h>
  45 #include <linux/times.h>
  46 #include <linux/limits.h>
  47 #include <linux/dcache.h>
  48 #include <linux/dnotify.h>
  49 #include <linux/syscalls.h>
  50 #include <linux/vmstat.h>
  51 #include <linux/nfs_fs.h>
  52 #include <linux/acpi.h>
  53 #include <linux/reboot.h>
  54 #include <linux/ftrace.h>
  55 #include <linux/perf_event.h>
  56 #include <linux/kprobes.h>
  57 #include <linux/pipe_fs_i.h>
  58 #include <linux/oom.h>
  59
  60 #include <asm/uaccess.h>
  61 #include <asm/processor.h>
  62
  63 #ifdef CONFIG_X86
  64 #include <asm/nmi.h>
  65 #include <asm/stacktrace.h>
  66 #include <asm/io.h>
  67 #endif
  68 #ifdef CONFIG_BSD_PROCESS_ACCT
  69 #include <linux/acct.h>
  70 #endif
  71 #ifdef CONFIG_RT_MUTEXES
  72 #include <linux/rtmutex.h>
  73 #endif
  74 #if defined(CONFIG_PROVE_LOCKING) || defined(CONFIG_LOCK_STAT)
  75 #include <linux/lockdep.h>
  76 #endif
  77 #ifdef CONFIG_CHR_DEV_SG
  78 #include <scsi/sg.h>
  79 #endif
  80
  81 #ifdef CONFIG_LOCKUP_DETECTOR
  82 #include <linux/nmi.h>
  83 #endif
  84
  85
  86 #if defined(CONFIG_SYSCTL)
  87
  88 /* External variables not in a header file. */
  89 extern int sysctl_overcommit_memory;
  90 extern int sysctl_overcommit_ratio;
  91 extern int max_threads;
  92 extern int core_uses_pid;
  93 extern int suid_dumpable;
  94 extern char core_pattern[];
  95 extern unsigned int core_pipe_limit;
  96 extern int pid_max;
  97 extern int min_free_kbytes;
  98 extern int pid_max_min, pid_max_max;
  99 extern int sysctl_drop_caches;
 100 extern int percpu_pagelist_fraction;
 101 extern int compat_log;
 102 extern int latencytop_enabled;
 103 extern int sysctl_nr_open_min, sysctl_nr_open_max;
 104 #ifndef CONFIG_MMU
 105 extern int sysctl_nr_trim_pages;
 106 #endif
 107 #ifdef CONFIG_BLOCK
 108 extern int blk_iopoll_enabled;
 109 #endif
 110
 111 /* Constants used for minimum and  maximum */
 112 #ifdef CONFIG_LOCKUP_DETECTOR
 113 static int sixty = 60;
 114 static int neg_one = -1;
 115 #endif
 116
 117 static int zero;
 118 static int __maybe_unused one = 1;
 119 static int __maybe_unused two = 2;
 120 static unsigned long one_ul = 1;
 121 static int one_hundred = 100;
 122 #ifdef CONFIG_PRINTK
 123 static int ten_thousand = 10000;
 124 #endif
 125
 126 /* this is needed for the proc_doulongvec_minmax of vm_dirty_bytes */
 127 static unsigned long dirty_bytes_min = 2 * PAGE_SIZE;
 128
 129 /* this is needed for the proc_dointvec_minmax for [fs_]overflow UID and GID */
 130 static int maxolduid = 65535;
 131 static int minolduid;
 132 static int min_percpu_pagelist_fract = 8;
 133
 134 static int ngroups_max = NGROUPS_MAX;
 135
 136 #ifdef CONFIG_INOTIFY_USER
 137 #include <linux/inotify.h>
 138 #endif
 139 #ifdef CONFIG_SPARC
 140 #include <asm/system.h>
 141 #endif
 142
 143 #ifdef CONFIG_SPARC64
 144 extern int sysctl_tsb_ratio;
 145 #endif
 146
 147 #ifdef __hppa__
 148 extern int pwrsw_enabled;
 149 extern int unaligned_enabled;
 150 #endif
 151
 152 #ifdef CONFIG_S390
 153 #ifdef CONFIG_MATHEMU
 154 extern int sysctl_ieee_emulation_warnings;
 155 #endif
 156 extern int sysctl_userprocess_debug;
 157 extern int spin_retry;
 158 #endif
 159
 160 #ifdef CONFIG_IA64
 161 extern int no_unaligned_warning;
 162 extern int unaligned_dump_stack;
 163 #endif
 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_EPOLL
 214 extern struct ctl_table epoll_table[];
 215 #endif
 216
 217 #ifdef HAVE_ARCH_PICK_MMAP_LAYOUT
 218 int sysctl_legacy_va_layout;
 219 #endif
 220
 221 /* The default sysctl tables: */
 222
 223 static struct ctl_table root_table[] = {
 224         {
 225                 .procname       = "kernel",
 226                 .mode           = 0555,
 227                 .child          = kern_table,
 228         },
 229         {
 230                 .procname       = "vm",
 231                 .mode           = 0555,
 232                 .child          = vm_table,
 233         },
 234         {
 235                 .procname       = "fs",
 236                 .mode           = 0555,
 237                 .child          = fs_table,
 238         },
 239         {
 240                 .procname       = "debug",
 241                 .mode           = 0555,
 242                 .child          = debug_table,
 243         },
 244         {
 245                 .procname       = "dev",
 246                 .mode           = 0555,
 247                 .child          = dev_table,
 248         },
 249 /*
 250  * NOTE: do not add new entries to this table unless you have read
 251  * Documentation/sysctl/ctl_unnumbered.txt
 252  */
 253         { }
 254 };
 255
 256 #ifdef CONFIG_SCHED_DEBUG
 257 static int min_sched_granularity_ns = 100000;           /* 100 usecs */
 258 static int max_sched_granularity_ns = NSEC_PER_SEC;     /* 1 second */
 259 static int min_wakeup_granularity_ns;                   /* 0 usecs */
 260 static int max_wakeup_granularity_ns = NSEC_PER_SEC;    /* 1 second */
 261 static int min_sched_tunable_scaling = SCHED_TUNABLESCALING_NONE;
 262 static int max_sched_tunable_scaling = SCHED_TUNABLESCALING_END-1;
 263 #endif
 264
 265 #ifdef CONFIG_COMPACTION
 266 static int min_extfrag_threshold;
 267 static int max_extfrag_threshold = 1000;
 268 #endif
 269
 270 static struct ctl_table kern_table[] = {
 271         {
 272                 .procname       = "sched_child_runs_first",
 273                 .data           = &sysctl_sched_child_runs_first,
 274                 .maxlen         = sizeof(unsigned int),
 275                 .mode           = 0644,
 276                 .proc_handler   = proc_dointvec,
 277         },
 278 #ifdef CONFIG_SCHED_DEBUG
 279         {
 280                 .procname       = "sched_min_granularity_ns",
 281                 .data           = &sysctl_sched_min_granularity,
 282                 .maxlen         = sizeof(unsigned int),
 283                 .mode           = 0644,
 284                 .proc_handler   = sched_proc_update_handler,
 285                 .extra1         = &min_sched_granularity_ns,
 286                 .extra2         = &max_sched_granularity_ns,
 287         },
 288         {
 289                 .procname       = "sched_latency_ns",
 290                 .data           = &sysctl_sched_latency,
 291                 .maxlen         = sizeof(unsigned int),
 292                 .mode           = 0644,
 293                 .proc_handler   = sched_proc_update_handler,
 294                 .extra1         = &min_sched_granularity_ns,
 295                 .extra2         = &max_sched_granularity_ns,
 296         },
 297         {
 298                 .procname       = "sched_wakeup_granularity_ns",
 299                 .data           = &sysctl_sched_wakeup_granularity,
 300                 .maxlen         = sizeof(unsigned int),
 301                 .mode           = 0644,
 302                 .proc_handler   = sched_proc_update_handler,
 303                 .extra1         = &min_wakeup_granularity_ns,
 304                 .extra2         = &max_wakeup_granularity_ns,
 305         },
 306         {
 307                 .procname       = "sched_tunable_scaling",
 308                 .data           = &sysctl_sched_tunable_scaling,
 309                 .maxlen         = sizeof(enum sched_tunable_scaling),
 310                 .mode           = 0644,
 311                 .proc_handler   = sched_proc_update_handler,
 312                 .extra1         = &min_sched_tunable_scaling,
 313                 .extra2         = &max_sched_tunable_scaling,
 314         },
 315         {
 316                 .procname       = "sched_migration_cost",
 317                 .data           = &sysctl_sched_migration_cost,
 318                 .maxlen         = sizeof(unsigned int),
 319                 .mode           = 0644,
 320                 .proc_handler   = proc_dointvec,
 321         },
 322         {
 323                 .procname       = "sched_nr_migrate",
 324                 .data           = &sysctl_sched_nr_migrate,
 325                 .maxlen         = sizeof(unsigned int),
 326                 .mode           = 0644,
 327                 .proc_handler   = proc_dointvec,
 328         },
 329         {
 330                 .procname       = "sched_time_avg",
 331                 .data           = &sysctl_sched_time_avg,
 332                 .maxlen         = sizeof(unsigned int),
 333                 .mode           = 0644,
 334                 .proc_handler   = proc_dointvec,
 335         },
 336         {
 337                 .procname       = "sched_shares_window",
 338                 .data           = &sysctl_sched_shares_window,
 339                 .maxlen         = sizeof(unsigned int),
 340                 .mode           = 0644,
 341                 .proc_handler   = proc_dointvec,
 342         },
 343         {
 344                 .procname       = "timer_migration",
 345                 .data           = &sysctl_timer_migration,
 346                 .maxlen         = sizeof(unsigned int),
 347                 .mode           = 0644,
 348                 .proc_handler   = proc_dointvec_minmax,
 349                 .extra1         = &zero,
 350                 .extra2         = &one,
 351         },
 352 #endif
 353         {
 354                 .procname       = "sched_rt_period_us",
 355                 .data           = &sysctl_sched_rt_period,
 356                 .maxlen         = sizeof(unsigned int),
 357                 .mode           = 0644,
 358                 .proc_handler   = sched_rt_handler,
 359         },
 360         {
 361                 .procname       = "sched_rt_runtime_us",
 362                 .data           = &sysctl_sched_rt_runtime,
 363                 .maxlen         = sizeof(int),
 364                 .mode           = 0644,
 365                 .proc_handler   = sched_rt_handler,
 366         },
 367         {
 368                 .procname       = "sched_compat_yield",
 369                 .data           = &sysctl_sched_compat_yield,
 370                 .maxlen         = sizeof(unsigned int),
 371                 .mode           = 0644,
 372                 .proc_handler   = proc_dointvec,
 373         },
 374 #ifdef CONFIG_SCHED_AUTOGROUP
 375         {
 376                 .procname       = "sched_autogroup_enabled",
 377                 .data           = &sysctl_sched_autogroup_enabled,
 378                 .maxlen         = sizeof(unsigned int),
 379                 .mode           = 0644,
 380                 .proc_handler   = proc_dointvec,
 381                 .extra1         = &zero,
 382                 .extra2         = &one,
 383         },
 384 #endif
 385 #ifdef CONFIG_PROVE_LOCKING
 386         {
 387                 .procname       = "prove_locking",
 388                 .data           = &prove_locking,
 389                 .maxlen         = sizeof(int),
 390                 .mode           = 0644,
 391                 .proc_handler   = proc_dointvec,
 392         },
 393 #endif
 394 #ifdef CONFIG_LOCK_STAT
 395         {
 396                 .procname       = "lock_stat",
 397                 .data           = &lock_stat,
 398                 .maxlen         = sizeof(int),
 399                 .mode           = 0644,
 400                 .proc_handler   = proc_dointvec,
 401         },
 402 #endif
 403         {
 404                 .procname       = "panic",
 405                 .data           = &panic_timeout,
 406                 .maxlen         = sizeof(int),
 407                 .mode           = 0644,
 408                 .proc_handler   = proc_dointvec,
 409         },
 410         {
 411                 .procname       = "core_uses_pid",
 412                 .data           = &core_uses_pid,
 413                 .maxlen         = sizeof(int),
 414                 .mode           = 0644,
 415                 .proc_handler   = proc_dointvec,
 416         },
 417         {
 418                 .procname       = "core_pattern",
 419                 .data           = core_pattern,
 420                 .maxlen         = CORENAME_MAX_SIZE,
 421                 .mode           = 0644,
 422                 .proc_handler   = proc_dostring,
 423         },
 424         {
 425                 .procname       = "core_pipe_limit",
 426                 .data           = &core_pipe_limit,
 427                 .maxlen         = sizeof(unsigned int),
 428                 .mode           = 0644,
 429                 .proc_handler   = proc_dointvec,
 430         },
 431 #ifdef CONFIG_PROC_SYSCTL
 432         {
 433                 .procname       = "tainted",
 434                 .maxlen         = sizeof(long),
 435                 .mode           = 0644,
 436                 .proc_handler   = proc_taint,
 437         },
 438 #endif
 439 #ifdef CONFIG_LATENCYTOP
 440         {
 441                 .procname       = "latencytop",
 442                 .data           = &latencytop_enabled,
 443                 .maxlen         = sizeof(int),
 444                 .mode           = 0644,
 445                 .proc_handler   = proc_dointvec,
 446         },
 447 #endif
 448 #ifdef CONFIG_BLK_DEV_INITRD
 449         {
 450                 .procname       = "real-root-dev",
 451                 .data           = &real_root_dev,
 452                 .maxlen         = sizeof(int),
 453                 .mode           = 0644,
 454                 .proc_handler   = proc_dointvec,
 455         },
 456 #endif
 457         {
 458                 .procname       = "print-fatal-signals",
 459                 .data           = &print_fatal_signals,
 460                 .maxlen         = sizeof(int),
 461                 .mode           = 0644,
 462                 .proc_handler   = proc_dointvec,
 463         },
 464 #ifdef CONFIG_SPARC
 465         {
 466                 .procname       = "reboot-cmd",
 467                 .data           = reboot_command,
 468                 .maxlen         = 256,
 469                 .mode           = 0644,
 470                 .proc_handler   = proc_dostring,
 471         },
 472         {
 473                 .procname       = "stop-a",
 474                 .data           = &stop_a_enabled,
 475                 .maxlen         = sizeof (int),
 476                 .mode           = 0644,
 477                 .proc_handler   = proc_dointvec,
 478         },
 479         {
 480                 .procname       = "scons-poweroff",
 481                 .data           = &scons_pwroff,
 482                 .maxlen         = sizeof (int),
 483                 .mode           = 0644,
 484                 .proc_handler   = proc_dointvec,
 485         },
 486 #endif
 487 #ifdef CONFIG_SPARC64
 488         {
 489                 .procname       = "tsb-ratio",
 490                 .data           = &sysctl_tsb_ratio,
 491                 .maxlen         = sizeof (int),
 492                 .mode           = 0644,
 493                 .proc_handler   = proc_dointvec,
 494         },
 495 #endif
 496 #ifdef __hppa__
 497         {
 498                 .procname       = "soft-power",
 499                 .data           = &pwrsw_enabled,
 500                 .maxlen         = sizeof (int),
 501                 .mode           = 0644,
 502                 .proc_handler   = proc_dointvec,
 503         },
 504         {
 505                 .procname       = "unaligned-trap",
 506                 .data           = &unaligned_enabled,
 507                 .maxlen         = sizeof (int),
 508                 .mode           = 0644,
 509                 .proc_handler   = proc_dointvec,
 510         },
 511 #endif
 512         {
 513                 .procname       = "ctrl-alt-del",
 514                 .data           = &C_A_D,
 515                 .maxlen         = sizeof(int),
 516                 .mode           = 0644,
 517                 .proc_handler   = proc_dointvec,
 518         },
 519 #ifdef CONFIG_FUNCTION_TRACER
 520         {
 521                 .procname       = "ftrace_enabled",
 522                 .data           = &ftrace_enabled,
 523                 .maxlen         = sizeof(int),
 524                 .mode           = 0644,
 525                 .proc_handler   = ftrace_enable_sysctl,
 526         },
 527 #endif
 528 #ifdef CONFIG_STACK_TRACER
 529         {
 530                 .procname       = "stack_tracer_enabled",
 531                 .data           = &stack_tracer_enabled,
 532                 .maxlen         = sizeof(int),
 533                 .mode           = 0644,
 534                 .proc_handler   = stack_trace_sysctl,
 535         },
 536 #endif
 537 #ifdef CONFIG_TRACING
 538         {
 539                 .procname       = "ftrace_dump_on_oops",
 540                 .data           = &ftrace_dump_on_oops,
 541                 .maxlen         = sizeof(int),
 542                 .mode           = 0644,
 543                 .proc_handler   = proc_dointvec,
 544         },
 545 #endif
 546 #ifdef CONFIG_MODULES
 547         {
 548                 .procname       = "modprobe",
 549                 .data           = &modprobe_path,
 550                 .maxlen         = KMOD_PATH_LEN,
 551                 .mode           = 0644,
 552                 .proc_handler   = proc_dostring,
 553         },
 554         {
 555                 .procname       = "modules_disabled",
 556                 .data           = &modules_disabled,
 557                 .maxlen         = sizeof(int),
 558                 .mode           = 0644,
 559                 /* only handle a transition from default "0" to "1" */
 560                 .proc_handler   = proc_dointvec_minmax,
 561                 .extra1         = &one,
 562                 .extra2         = &one,
 563         },
 564 #endif
 565 #ifdef CONFIG_HOTPLUG
 566         {
 567                 .procname       = "hotplug",
 568                 .data           = &uevent_helper,
 569                 .maxlen         = UEVENT_HELPER_PATH_LEN,
 570                 .mode           = 0644,
 571                 .proc_handler   = proc_dostring,
 572         },
 573 #endif
 574 #ifdef CONFIG_CHR_DEV_SG
 575         {
 576                 .procname       = "sg-big-buff",
 577                 .data           = &sg_big_buff,
 578                 .maxlen         = sizeof (int),
 579                 .mode           = 0444,
 580                 .proc_handler   = proc_dointvec,
 581         },
 582 #endif
 583 #ifdef CONFIG_BSD_PROCESS_ACCT
 584         {
 585                 .procname       = "acct",
 586                 .data           = &acct_parm,
 587                 .maxlen         = 3*sizeof(int),
 588                 .mode           = 0644,
 589                 .proc_handler   = proc_dointvec,
 590         },
 591 #endif
 592 #ifdef CONFIG_MAGIC_SYSRQ
 593         {
 594                 .procname       = "sysrq",
 595                 .data           = &__sysrq_enabled,
 596                 .maxlen         = sizeof (int),
 597                 .mode           = 0644,
 598                 .proc_handler   = sysrq_sysctl_handler,
 599         },
 600 #endif
 601 #ifdef CONFIG_PROC_SYSCTL
 602         {
 603                 .procname       = "cad_pid",
 604                 .data           = NULL,
 605                 .maxlen         = sizeof (int),
 606                 .mode           = 0600,
 607                 .proc_handler   = proc_do_cad_pid,
 608         },
 609 #endif
 610         {
 611                 .procname       = "threads-max",
 612                 .data           = &max_threads,
 613                 .maxlen         = sizeof(int),
 614                 .mode           = 0644,
 615                 .proc_handler   = proc_dointvec,
 616         },
 617         {
 618                 .procname       = "random",
 619                 .mode           = 0555,
 620                 .child          = random_table,
 621         },
 622         {
 623                 .procname       = "overflowuid",
 624                 .data           = &overflowuid,
 625                 .maxlen         = sizeof(int),
 626                 .mode           = 0644,
 627                 .proc_handler   = proc_dointvec_minmax,
 628                 .extra1         = &minolduid,
 629                 .extra2         = &maxolduid,
 630         },
 631         {
 632                 .procname       = "overflowgid",
 633                 .data           = &overflowgid,
 634                 .maxlen         = sizeof(int),
 635                 .mode           = 0644,
 636                 .proc_handler   = proc_dointvec_minmax,
 637                 .extra1         = &minolduid,
 638                 .extra2         = &maxolduid,
 639         },
 640 #ifdef CONFIG_S390
 641 #ifdef CONFIG_MATHEMU
 642         {
 643                 .procname       = "ieee_emulation_warnings",
 644                 .data           = &sysctl_ieee_emulation_warnings,
 645                 .maxlen         = sizeof(int),
 646                 .mode           = 0644,
 647                 .proc_handler   = proc_dointvec,
 648         },
 649 #endif
 650         {
 651                 .procname       = "userprocess_debug",
 652                 .data           = &show_unhandled_signals,
 653                 .maxlen         = sizeof(int),
 654                 .mode           = 0644,
 655                 .proc_handler   = proc_dointvec,
 656         },
 657 #endif
 658         {
 659                 .procname       = "pid_max",
 660                 .data           = &pid_max,
 661                 .maxlen         = sizeof (int),
 662                 .mode           = 0644,
 663                 .proc_handler   = proc_dointvec_minmax,
 664                 .extra1         = &pid_max_min,
 665                 .extra2         = &pid_max_max,
 666         },
 667         {
 668                 .procname       = "panic_on_oops",
 669                 .data           = &panic_on_oops,
 670                 .maxlen         = sizeof(int),
 671                 .mode           = 0644,
 672                 .proc_handler   = proc_dointvec,
 673         },
 674 #if defined CONFIG_PRINTK
 675         {
 676                 .procname       = "printk",
 677                 .data           = &console_loglevel,
 678                 .maxlen         = 4*sizeof(int),
 679                 .mode           = 0644,
 680                 .proc_handler   = proc_dointvec,
 681         },
 682         {
 683                 .procname       = "printk_ratelimit",
 684                 .data           = &printk_ratelimit_state.interval,
 685                 .maxlen         = sizeof(int),
 686                 .mode           = 0644,
 687                 .proc_handler   = proc_dointvec_jiffies,
 688         },
 689         {
 690                 .procname       = "printk_ratelimit_burst",
 691                 .data           = &printk_ratelimit_state.burst,
 692                 .maxlen         = sizeof(int),
 693                 .mode           = 0644,
 694                 .proc_handler   = proc_dointvec,
 695         },
 696         {
 697                 .procname       = "printk_delay",
 698                 .data           = &printk_delay_msec,
 699                 .maxlen         = sizeof(int),
 700                 .mode           = 0644,
 701                 .proc_handler   = proc_dointvec_minmax,
 702                 .extra1         = &zero,
 703                 .extra2         = &ten_thousand,
 704         },
 705         {
 706                 .procname       = "dmesg_restrict",
 707                 .data           = &dmesg_restrict,
 708                 .maxlen         = sizeof(int),
 709                 .mode           = 0644,
 710                 .proc_handler   = proc_dointvec_minmax,
 711                 .extra1         = &zero,
 712                 .extra2         = &one,
 713         },
 714         {
 715                 .procname       = "kptr_restrict",
 716                 .data           = &kptr_restrict,
 717                 .maxlen         = sizeof(int),
 718                 .mode           = 0644,
 719                 .proc_handler   = proc_dointvec_minmax,
 720                 .extra1         = &zero,
 721                 .extra2         = &two,
 722         },
 723 #endif
 724         {
 725                 .procname       = "ngroups_max",
 726                 .data           = &ngroups_max,
 727                 .maxlen         = sizeof (int),
 728                 .mode           = 0444,
 729                 .proc_handler   = proc_dointvec,
 730         },
 731 #if defined(CONFIG_LOCKUP_DETECTOR)
 732         {
 733                 .procname       = "watchdog",
 734                 .data           = &watchdog_enabled,
 735                 .maxlen         = sizeof (int),
 736                 .mode           = 0644,
 737                 .proc_handler   = proc_dowatchdog_enabled,
 738         },
 739         {
 740                 .procname       = "watchdog_thresh",
 741                 .data           = &softlockup_thresh,
 742                 .maxlen         = sizeof(int),
 743                 .mode           = 0644,
 744                 .proc_handler   = proc_dowatchdog_thresh,
 745                 .extra1         = &neg_one,
 746                 .extra2         = &sixty,
 747         },
 748         {
 749                 .procname       = "softlockup_panic",
 750                 .data           = &softlockup_panic,
 751                 .maxlen         = sizeof(int),
 752                 .mode           = 0644,
 753                 .proc_handler   = proc_dointvec_minmax,
 754                 .extra1         = &zero,
 755                 .extra2         = &one,
 756         },
 757         {
 758                 .procname       = "nmi_watchdog",
 759                 .data           = &watchdog_enabled,
 760                 .maxlen         = sizeof (int),
 761                 .mode           = 0644,
 762                 .proc_handler   = proc_dowatchdog_enabled,
 763         },
 764 #endif
 765 #if defined(CONFIG_X86_LOCAL_APIC) && defined(CONFIG_X86)
 766         {
 767                 .procname       = "unknown_nmi_panic",
 768                 .data           = &unknown_nmi_panic,
 769                 .maxlen         = sizeof (int),
 770                 .mode           = 0644,
 771                 .proc_handler   = proc_dointvec,
 772         },
 773 #endif
 774 #if defined(CONFIG_X86)
 775         {
 776                 .procname       = "panic_on_unrecovered_nmi",
 777                 .data           = &panic_on_unrecovered_nmi,
 778                 .maxlen         = sizeof(int),
 779                 .mode           = 0644,
 780                 .proc_handler   = proc_dointvec,
 781         },
 782         {
 783                 .procname       = "panic_on_io_nmi",
 784                 .data           = &panic_on_io_nmi,
 785                 .maxlen         = sizeof(int),
 786                 .mode           = 0644,
 787                 .proc_handler   = proc_dointvec,
 788         },
 789         {
 790                 .procname       = "bootloader_type",
 791                 .data           = &bootloader_type,
 792                 .maxlen         = sizeof (int),
 793                 .mode           = 0444,
 794                 .proc_handler   = proc_dointvec,
 795         },
 796         {
 797                 .procname       = "bootloader_version",
 798                 .data           = &bootloader_version,
 799                 .maxlen         = sizeof (int),
 800                 .mode           = 0444,
 801                 .proc_handler   = proc_dointvec,
 802         },
 803         {
 804                 .procname       = "kstack_depth_to_print",
 805                 .data           = &kstack_depth_to_print,
 806                 .maxlen         = sizeof(int),
 807                 .mode           = 0644,
 808                 .proc_handler   = proc_dointvec,
 809         },
 810         {
 811                 .procname       = "io_delay_type",
 812                 .data           = &io_delay_type,
 813                 .maxlen         = sizeof(int),
 814                 .mode           = 0644,
 815                 .proc_handler   = proc_dointvec,
 816         },
 817 #endif
 818 #if defined(CONFIG_MMU)
 819         {
 820                 .procname       = "randomize_va_space",
 821                 .data           = &randomize_va_space,
 822                 .maxlen         = sizeof(int),
 823                 .mode           = 0644,
 824                 .proc_handler   = proc_dointvec,
 825         },
 826 #endif
 827 #if defined(CONFIG_S390) && defined(CONFIG_SMP)
 828         {
 829                 .procname       = "spin_retry",
 830                 .data           = &spin_retry,
 831                 .maxlen         = sizeof (int),
 832                 .mode           = 0644,
 833                 .proc_handler   = proc_dointvec,
 834         },
 835 #endif
 836 #if     defined(CONFIG_ACPI_SLEEP) && defined(CONFIG_X86)
 837         {
 838                 .procname       = "acpi_video_flags",
 839                 .data           = &acpi_realmode_flags,
 840                 .maxlen         = sizeof (unsigned long),
 841                 .mode           = 0644,
 842                 .proc_handler   = proc_doulongvec_minmax,
 843         },
 844 #endif
 845 #ifdef CONFIG_IA64
 846         {
 847                 .procname       = "ignore-unaligned-usertrap",
 848                 .data           = &no_unaligned_warning,
 849                 .maxlen         = sizeof (int),
 850                 .mode           = 0644,
 851                 .proc_handler   = proc_dointvec,
 852         },
 853         {
 854                 .procname       = "unaligned-dump-stack",
 855                 .data           = &unaligned_dump_stack,
 856                 .maxlen         = sizeof (int),
 857                 .mode           = 0644,
 858                 .proc_handler   = proc_dointvec,
 859         },
 860 #endif
 861 #ifdef CONFIG_DETECT_HUNG_TASK
 862         {
 863                 .procname       = "hung_task_panic",
 864                 .data           = &sysctl_hung_task_panic,
 865                 .maxlen         = sizeof(int),
 866                 .mode           = 0644,
 867                 .proc_handler   = proc_dointvec_minmax,
 868                 .extra1         = &zero,
 869                 .extra2         = &one,
 870         },
 871         {
 872                 .procname       = "hung_task_check_count",
 873                 .data           = &sysctl_hung_task_check_count,
 874                 .maxlen         = sizeof(unsigned long),
 875                 .mode           = 0644,
 876                 .proc_handler   = proc_doulongvec_minmax,
 877         },
 878         {
 879                 .procname       = "hung_task_timeout_secs",
 880                 .data           = &sysctl_hung_task_timeout_secs,
 881                 .maxlen         = sizeof(unsigned long),
 882                 .mode           = 0644,
 883                 .proc_handler   = proc_dohung_task_timeout_secs,
 884         },
 885         {
 886                 .procname       = "hung_task_warnings",
 887                 .data           = &sysctl_hung_task_warnings,
 888                 .maxlen         = sizeof(unsigned long),
 889                 .mode           = 0644,
 890                 .proc_handler   = proc_doulongvec_minmax,
 891         },
 892 #endif
 893 #ifdef CONFIG_COMPAT
 894         {
 895                 .procname       = "compat-log",
 896                 .data           = &compat_log,
 897                 .maxlen         = sizeof (int),
 898                 .mode           = 0644,
 899                 .proc_handler   = proc_dointvec,
 900         },
 901 #endif
 902 #ifdef CONFIG_RT_MUTEXES
 903         {
 904                 .procname       = "max_lock_depth",
 905                 .data           = &max_lock_depth,
 906                 .maxlen         = sizeof(int),
 907                 .mode           = 0644,
 908                 .proc_handler   = proc_dointvec,
 909         },
 910 #endif
 911         {
 912                 .procname       = "poweroff_cmd",
 913                 .data           = &poweroff_cmd,
 914                 .maxlen         = POWEROFF_CMD_PATH_LEN,
 915                 .mode           = 0644,
 916                 .proc_handler   = proc_dostring,
 917         },
 918 #ifdef CONFIG_KEYS
 919         {
 920                 .procname       = "keys",
 921                 .mode           = 0555,
 922                 .child          = key_sysctls,
 923         },
 924 #endif
 925 #ifdef CONFIG_RCU_TORTURE_TEST
 926         {
 927                 .procname       = "rcutorture_runnable",
 928                 .data           = &rcutorture_runnable,
 929                 .maxlen         = sizeof(int),
 930                 .mode           = 0644,
 931                 .proc_handler   = proc_dointvec,
 932         },
 933 #endif
 934 #ifdef CONFIG_PERF_EVENTS
 935         {
 936                 .procname       = "perf_event_paranoid",
 937                 .data           = &sysctl_perf_event_paranoid,
 938                 .maxlen         = sizeof(sysctl_perf_event_paranoid),
 939                 .mode           = 0644,
 940                 .proc_handler   = proc_dointvec,
 941         },
 942         {
 943                 .procname       = "perf_event_mlock_kb",
 944                 .data           = &sysctl_perf_event_mlock,
 945                 .maxlen         = sizeof(sysctl_perf_event_mlock),
 946                 .mode           = 0644,
 947                 .proc_handler   = proc_dointvec,
 948         },
 949         {
 950                 .procname       = "perf_event_max_sample_rate",
 951                 .data           = &sysctl_perf_event_sample_rate,
 952                 .maxlen         = sizeof(sysctl_perf_event_sample_rate),
 953                 .mode           = 0644,
 954                 .proc_handler   = proc_dointvec,
 955         },
 956 #endif
 957 #ifdef CONFIG_KMEMCHECK
 958         {
 959                 .procname       = "kmemcheck",
 960                 .data           = &kmemcheck_enabled,
 961                 .maxlen         = sizeof(int),
 962                 .mode           = 0644,
 963                 .proc_handler   = proc_dointvec,
 964         },
 965 #endif
 966 #ifdef CONFIG_BLOCK
 967         {
 968                 .procname       = "blk_iopoll",
 969                 .data           = &blk_iopoll_enabled,
 970                 .maxlen         = sizeof(int),
 971                 .mode           = 0644,
 972                 .proc_handler   = proc_dointvec,
 973         },
 974 #endif
 975 /*
 976  * NOTE: do not add new entries to this table unless you have read
 977  * Documentation/sysctl/ctl_unnumbered.txt
 978  */
 979         { }
 980 };
 981
 982 static struct ctl_table vm_table[] = {
 983         {
 984                 .procname       = "overcommit_memory",
 985                 .data           = &sysctl_overcommit_memory,
 986                 .maxlen         = sizeof(sysctl_overcommit_memory),
 987                 .mode           = 0644,
 988                 .proc_handler   = proc_dointvec,
 989         },
 990         {
 991                 .procname       = "panic_on_oom",
 992                 .data           = &sysctl_panic_on_oom,
 993                 .maxlen         = sizeof(sysctl_panic_on_oom),
 994                 .mode           = 0644,
 995                 .proc_handler   = proc_dointvec,
 996         },
 997         {
 998                 .procname       = "oom_kill_allocating_task",
 999                 .data           = &sysctl_oom_kill_allocating_task,
1000                 .maxlen         = sizeof(sysctl_oom_kill_allocating_task),
1001                 .mode           = 0644,
1002                 .proc_handler   = proc_dointvec,
1003         },
1004         {
1005                 .procname       = "oom_dump_tasks",
1006                 .data           = &sysctl_oom_dump_tasks,
1007                 .maxlen         = sizeof(sysctl_oom_dump_tasks),
1008                 .mode           = 0644,
1009                 .proc_handler   = proc_dointvec,
1010         },
1011         {
1012                 .procname       = "overcommit_ratio",
1013                 .data           = &sysctl_overcommit_ratio,
1014                 .maxlen         = sizeof(sysctl_overcommit_ratio),
1015                 .mode           = 0644,
1016                 .proc_handler   = proc_dointvec,
1017         },
1018         {
1019                 .procname       = "page-cluster",
1020                 .data           = &page_cluster,
1021                 .maxlen         = sizeof(int),
1022                 .mode           = 0644,
1023                 .proc_handler   = proc_dointvec,
1024         },
1025         {
1026                 .procname       = "dirty_background_ratio",
1027                 .data           = &dirty_background_ratio,
1028                 .maxlen         = sizeof(dirty_background_ratio),
1029                 .mode           = 0644,
1030                 .proc_handler   = dirty_background_ratio_handler,
1031                 .extra1         = &zero,
1032                 .extra2         = &one_hundred,
1033         },
1034         {
1035                 .procname       = "dirty_background_bytes",
1036                 .data           = &dirty_background_bytes,
1037                 .maxlen         = sizeof(dirty_background_bytes),
1038                 .mode           = 0644,
1039                 .proc_handler   = dirty_background_bytes_handler,
1040                 .extra1         = &one_ul,
1041         },
1042         {
1043                 .procname       = "dirty_ratio",
1044                 .data           = &vm_dirty_ratio,
1045                 .maxlen         = sizeof(vm_dirty_ratio),
1046                 .mode           = 0644,
1047                 .proc_handler   = dirty_ratio_handler,
1048                 .extra1         = &zero,
1049                 .extra2         = &one_hundred,
1050         },
1051         {
1052                 .procname       = "dirty_bytes",
1053                 .data           = &vm_dirty_bytes,
1054                 .maxlen         = sizeof(vm_dirty_bytes),
1055                 .mode           = 0644,
1056                 .proc_handler   = dirty_bytes_handler,
1057                 .extra1         = &dirty_bytes_min,
1058         },
1059         {
1060                 .procname       = "dirty_writeback_centisecs",
1061                 .data           = &dirty_writeback_interval,
1062                 .maxlen         = sizeof(dirty_writeback_interval),
1063                 .mode           = 0644,
1064                 .proc_handler   = dirty_writeback_centisecs_handler,
1065         },
1066         {
1067                 .procname       = "dirty_expire_centisecs",
1068                 .data           = &dirty_expire_interval,
1069                 .maxlen         = sizeof(dirty_expire_interval),
1070                 .mode           = 0644,
1071                 .proc_handler   = proc_dointvec,
1072         },
1073         {
1074                 .procname       = "nr_pdflush_threads",
1075                 .data           = &nr_pdflush_threads,
1076                 .maxlen         = sizeof nr_pdflush_threads,
1077                 .mode           = 0444 /* read-only*/,
1078                 .proc_handler   = proc_dointvec,
1079         },
1080         {
1081                 .procname       = "swappiness",
1082                 .data           = &vm_swappiness,
1083                 .maxlen         = sizeof(vm_swappiness),
1084                 .mode           = 0644,
1085                 .proc_handler   = proc_dointvec_minmax,
1086                 .extra1         = &zero,
1087                 .extra2         = &one_hundred,
1088         },
1089 #ifdef CONFIG_HUGETLB_PAGE
1090         {
1091                 .procname       = "nr_hugepages",
1092                 .data           = NULL,
1093                 .maxlen         = sizeof(unsigned long),
1094                 .mode           = 0644,
1095                 .proc_handler   = hugetlb_sysctl_handler,
1096                 .extra1         = (void *)&hugetlb_zero,
1097                 .extra2         = (void *)&hugetlb_infinity,
1098         },
1099 #ifdef CONFIG_NUMA
1100         {
1101                 .procname       = "nr_hugepages_mempolicy",
1102                 .data           = NULL,
1103                 .maxlen         = sizeof(unsigned long),
1104                 .mode           = 0644,
1105                 .proc_handler   = &hugetlb_mempolicy_sysctl_handler,
1106                 .extra1         = (void *)&hugetlb_zero,
1107                 .extra2         = (void *)&hugetlb_infinity,
1108         },
1109 #endif
1110          {
1111                 .procname       = "hugetlb_shm_group",
1112                 .data           = &sysctl_hugetlb_shm_group,
1113                 .maxlen         = sizeof(gid_t),
1114                 .mode           = 0644,
1115                 .proc_handler   = proc_dointvec,
1116          },
1117          {
1118                 .procname       = "hugepages_treat_as_movable",
1119                 .data           = &hugepages_treat_as_movable,
1120                 .maxlen         = sizeof(int),
1121                 .mode           = 0644,
1122                 .proc_handler   = hugetlb_treat_movable_handler,
1123         },
1124         {
1125                 .procname       = "nr_overcommit_hugepages",
1126                 .data           = NULL,
1127                 .maxlen         = sizeof(unsigned long),
1128                 .mode           = 0644,
1129                 .proc_handler   = hugetlb_overcommit_handler,
1130                 .extra1         = (void *)&hugetlb_zero,
1131                 .extra2         = (void *)&hugetlb_infinity,
1132         },
1133 #endif
1134         {
1135                 .procname       = "lowmem_reserve_ratio",
1136                 .data           = &sysctl_lowmem_reserve_ratio,
1137                 .maxlen         = sizeof(sysctl_lowmem_reserve_ratio),
1138                 .mode           = 0644,
1139                 .proc_handler   = lowmem_reserve_ratio_sysctl_handler,
1140         },
1141         {
1142                 .procname       = "drop_caches",
1143                 .data           = &sysctl_drop_caches,
1144                 .maxlen         = sizeof(int),
1145                 .mode           = 0644,
1146                 .proc_handler   = drop_caches_sysctl_handler,
1147         },
1148 #ifdef CONFIG_COMPACTION
1149         {
1150                 .procname       = "compact_memory",
1151                 .data           = &sysctl_compact_memory,
1152                 .maxlen         = sizeof(int),
1153                 .mode           = 0200,
1154                 .proc_handler   = sysctl_compaction_handler,
1155         },
1156         {
1157                 .procname       = "extfrag_threshold",
1158                 .data           = &sysctl_extfrag_threshold,
1159                 .maxlen         = sizeof(int),
1160                 .mode           = 0644,
1161                 .proc_handler   = sysctl_extfrag_handler,
1162                 .extra1         = &min_extfrag_threshold,
1163                 .extra2         = &max_extfrag_threshold,
1164         },
1165
1166 #endif /* CONFIG_COMPACTION */
1167         {
1168                 .procname       = "min_free_kbytes",
1169                 .data           = &min_free_kbytes,
1170                 .maxlen         = sizeof(min_free_kbytes),
1171                 .mode           = 0644,
1172                 .proc_handler   = min_free_kbytes_sysctl_handler,
1173                 .extra1         = &zero,
1174         },
1175         {
1176                 .procname       = "percpu_pagelist_fraction",
1177                 .data           = &percpu_pagelist_fraction,
1178                 .maxlen         = sizeof(percpu_pagelist_fraction),
1179                 .mode           = 0644,
1180                 .proc_handler   = percpu_pagelist_fraction_sysctl_handler,
1181                 .extra1         = &min_percpu_pagelist_fract,
1182         },
1183 #ifdef CONFIG_MMU
1184         {
1185                 .procname       = "max_map_count",
1186                 .data           = &sysctl_max_map_count,
1187                 .maxlen         = sizeof(sysctl_max_map_count),
1188                 .mode           = 0644,
1189                 .proc_handler   = proc_dointvec_minmax,
1190                 .extra1         = &zero,
1191         },
1192 #else
1193         {
1194                 .procname       = "nr_trim_pages",
1195                 .data           = &sysctl_nr_trim_pages,
1196                 .maxlen         = sizeof(sysctl_nr_trim_pages),
1197                 .mode           = 0644,
1198                 .proc_handler   = proc_dointvec_minmax,
1199                 .extra1         = &zero,
1200         },
1201 #endif
1202         {
1203                 .procname       = "laptop_mode",
1204                 .data           = &laptop_mode,
1205                 .maxlen         = sizeof(laptop_mode),
1206                 .mode           = 0644,
1207                 .proc_handler   = proc_dointvec_jiffies,
1208         },
1209         {
1210                 .procname       = "block_dump",
1211                 .data           = &block_dump,
1212                 .maxlen         = sizeof(block_dump),
1213                 .mode           = 0644,
1214                 .proc_handler   = proc_dointvec,
1215                 .extra1         = &zero,
1216         },
1217         {
1218                 .procname       = "vfs_cache_pressure",
1219                 .data           = &sysctl_vfs_cache_pressure,
1220                 .maxlen         = sizeof(sysctl_vfs_cache_pressure),
1221                 .mode           = 0644,
1222                 .proc_handler   = proc_dointvec,
1223                 .extra1         = &zero,
1224         },
1225 #ifdef HAVE_ARCH_PICK_MMAP_LAYOUT
1226         {
1227                 .procname       = "legacy_va_layout",
1228                 .data           = &sysctl_legacy_va_layout,
1229                 .maxlen         = sizeof(sysctl_legacy_va_layout),
1230                 .mode           = 0644,
1231                 .proc_handler   = proc_dointvec,
1232                 .extra1         = &zero,
1233         },
1234 #endif
1235 #ifdef CONFIG_NUMA
1236         {
1237                 .procname       = "zone_reclaim_mode",
1238                 .data           = &zone_reclaim_mode,
1239                 .maxlen         = sizeof(zone_reclaim_mode),
1240                 .mode           = 0644,
1241                 .proc_handler   = proc_dointvec,
1242                 .extra1         = &zero,
1243         },
1244         {
1245                 .procname       = "min_unmapped_ratio",
1246                 .data           = &sysctl_min_unmapped_ratio,
1247                 .maxlen         = sizeof(sysctl_min_unmapped_ratio),
1248                 .mode           = 0644,
1249                 .proc_handler   = sysctl_min_unmapped_ratio_sysctl_handler,
1250                 .extra1         = &zero,
1251                 .extra2         = &one_hundred,
1252         },
1253         {
1254                 .procname       = "min_slab_ratio",
1255                 .data           = &sysctl_min_slab_ratio,
1256                 .maxlen         = sizeof(sysctl_min_slab_ratio),
1257                 .mode           = 0644,
1258                 .proc_handler   = sysctl_min_slab_ratio_sysctl_handler,
1259                 .extra1         = &zero,
1260                 .extra2         = &one_hundred,
1261         },
1262 #endif
1263 #ifdef CONFIG_SMP
1264         {
1265                 .procname       = "stat_interval",
1266                 .data           = &sysctl_stat_interval,
1267                 .maxlen         = sizeof(sysctl_stat_interval),
1268                 .mode           = 0644,
1269                 .proc_handler   = proc_dointvec_jiffies,
1270         },
1271 #endif
1272 #ifdef CONFIG_MMU
1273         {
1274                 .procname       = "mmap_min_addr",
1275                 .data           = &dac_mmap_min_addr,
1276                 .maxlen         = sizeof(unsigned long),
1277                 .mode           = 0644,
1278                 .proc_handler   = mmap_min_addr_handler,
1279         },
1280 #endif
1281 #ifdef CONFIG_NUMA
1282         {
1283                 .procname       = "numa_zonelist_order",
1284                 .data           = &numa_zonelist_order,
1285                 .maxlen         = NUMA_ZONELIST_ORDER_LEN,
1286                 .mode           = 0644,
1287                 .proc_handler   = numa_zonelist_order_handler,
1288         },
1289 #endif
1290 #if (defined(CONFIG_X86_32) && !defined(CONFIG_UML))|| \
1291    (defined(CONFIG_SUPERH) && defined(CONFIG_VSYSCALL))
1292         {
1293                 .procname       = "vdso_enabled",
1294                 .data           = &vdso_enabled,
1295                 .maxlen         = sizeof(vdso_enabled),
1296                 .mode           = 0644,
1297                 .proc_handler   = proc_dointvec,
1298                 .extra1         = &zero,
1299         },
1300 #endif
1301 #ifdef CONFIG_HIGHMEM
1302         {
1303                 .procname       = "highmem_is_dirtyable",
1304                 .data           = &vm_highmem_is_dirtyable,
1305                 .maxlen         = sizeof(vm_highmem_is_dirtyable),
1306                 .mode           = 0644,
1307                 .proc_handler   = proc_dointvec_minmax,
1308                 .extra1         = &zero,
1309                 .extra2         = &one,
1310         },
1311 #endif
1312         {
1313                 .procname       = "scan_unevictable_pages",
1314                 .data           = &scan_unevictable_pages,
1315                 .maxlen         = sizeof(scan_unevictable_pages),
1316                 .mode           = 0644,
1317                 .proc_handler   = scan_unevictable_handler,
1318         },
1319 #ifdef CONFIG_MEMORY_FAILURE
1320         {
1321                 .procname       = "memory_failure_early_kill",
1322                 .data           = &sysctl_memory_failure_early_kill,
1323                 .maxlen         = sizeof(sysctl_memory_failure_early_kill),
1324                 .mode           = 0644,
1325                 .proc_handler   = proc_dointvec_minmax,
1326                 .extra1         = &zero,
1327                 .extra2         = &one,
1328         },
1329         {
1330                 .procname       = "memory_failure_recovery",
1331                 .data           = &sysctl_memory_failure_recovery,
1332                 .maxlen         = sizeof(sysctl_memory_failure_recovery),
1333                 .mode           = 0644,
1334                 .proc_handler   = proc_dointvec_minmax,
1335                 .extra1         = &zero,
1336                 .extra2         = &one,
1337         },
1338 #endif
1339
1340 /*
1341  * NOTE: do not add new entries to this table unless you have read
1342  * Documentation/sysctl/ctl_unnumbered.txt
1343  */
1344         { }
1345 };
1346
1347 #if defined(CONFIG_BINFMT_MISC) || defined(CONFIG_BINFMT_MISC_MODULE)
1348 static struct ctl_table binfmt_misc_table[] = {
1349         { }
1350 };
1351 #endif
1352
1353 static struct ctl_table fs_table[] = {
1354         {
1355                 .procname       = "inode-nr",
1356                 .data           = &inodes_stat,
1357                 .maxlen         = 2*sizeof(int),
1358                 .mode           = 0444,
1359                 .proc_handler   = proc_nr_inodes,
1360         },
1361         {
1362                 .procname       = "inode-state",
1363                 .data           = &inodes_stat,
1364                 .maxlen         = 7*sizeof(int),
1365                 .mode           = 0444,
1366                 .proc_handler   = proc_nr_inodes,
1367         },
1368         {
1369                 .procname       = "file-nr",
1370                 .data           = &files_stat,
1371                 .maxlen         = sizeof(files_stat),
1372                 .mode           = 0444,
1373                 .proc_handler   = proc_nr_files,
1374         },
1375         {
1376                 .procname       = "file-max",
1377                 .data           = &files_stat.max_files,
1378                 .maxlen         = sizeof(files_stat.max_files),
1379                 .mode           = 0644,
1380                 .proc_handler   = proc_doulongvec_minmax,
1381         },
1382         {
1383                 .procname       = "nr_open",
1384                 .data           = &sysctl_nr_open,
1385                 .maxlen         = sizeof(int),
1386                 .mode           = 0644,
1387                 .proc_handler   = proc_dointvec_minmax,
1388                 .extra1         = &sysctl_nr_open_min,
1389                 .extra2         = &sysctl_nr_open_max,
1390         },
1391         {
1392                 .procname       = "dentry-state",
1393                 .data           = &dentry_stat,
1394                 .maxlen         = 6*sizeof(int),
1395                 .mode           = 0444,
1396                 .proc_handler   = proc_nr_dentry,
1397         },
1398         {
1399                 .procname       = "overflowuid",
1400                 .data           = &fs_overflowuid,
1401                 .maxlen         = sizeof(int),
1402                 .mode           = 0644,
1403                 .proc_handler   = proc_dointvec_minmax,
1404                 .extra1         = &minolduid,
1405                 .extra2         = &maxolduid,
1406         },
1407         {
1408                 .procname       = "overflowgid",
1409                 .data           = &fs_overflowgid,
1410                 .maxlen         = sizeof(int),
1411                 .mode           = 0644,
1412                 .proc_handler   = proc_dointvec_minmax,
1413                 .extra1         = &minolduid,
1414                 .extra2         = &maxolduid,
1415         },
1416 #ifdef CONFIG_FILE_LOCKING
1417         {
1418                 .procname       = "leases-enable",
1419                 .data           = &leases_enable,
1420                 .maxlen         = sizeof(int),
1421                 .mode           = 0644,
1422                 .proc_handler   = proc_dointvec,
1423         },
1424 #endif
1425 #ifdef CONFIG_DNOTIFY
1426         {
1427                 .procname       = "dir-notify-enable",
1428                 .data           = &dir_notify_enable,
1429                 .maxlen         = sizeof(int),
1430                 .mode           = 0644,
1431                 .proc_handler   = proc_dointvec,
1432         },
1433 #endif
1434 #ifdef CONFIG_MMU
1435 #ifdef CONFIG_FILE_LOCKING
1436         {
1437                 .procname       = "lease-break-time",
1438                 .data           = &lease_break_time,
1439                 .maxlen         = sizeof(int),
1440                 .mode           = 0644,
1441                 .proc_handler   = proc_dointvec,
1442         },
1443 #endif
1444 #ifdef CONFIG_AIO
1445         {
1446                 .procname       = "aio-nr",
1447                 .data           = &aio_nr,
1448                 .maxlen         = sizeof(aio_nr),
1449                 .mode           = 0444,
1450                 .proc_handler   = proc_doulongvec_minmax,
1451         },
1452         {
1453                 .procname       = "aio-max-nr",
1454                 .data           = &aio_max_nr,
1455                 .maxlen         = sizeof(aio_max_nr),
1456                 .mode           = 0644,
1457                 .proc_handler   = proc_doulongvec_minmax,
1458         },
1459 #endif /* CONFIG_AIO */
1460 #ifdef CONFIG_INOTIFY_USER
1461         {
1462                 .procname       = "inotify",
1463                 .mode           = 0555,
1464                 .child          = inotify_table,
1465         },
1466 #endif
1467 #ifdef CONFIG_EPOLL
1468         {
1469                 .procname       = "epoll",
1470                 .mode           = 0555,
1471                 .child          = epoll_table,
1472         },
1473 #endif
1474 #endif
1475         {
1476                 .procname       = "suid_dumpable",
1477                 .data           = &suid_dumpable,
1478                 .maxlen         = sizeof(int),
1479                 .mode           = 0644,
1480                 .proc_handler   = proc_dointvec_minmax,
1481                 .extra1         = &zero,
1482                 .extra2         = &two,
1483         },
1484 #if defined(CONFIG_BINFMT_MISC) || defined(CONFIG_BINFMT_MISC_MODULE)
1485         {
1486                 .procname       = "binfmt_misc",
1487                 .mode           = 0555,
1488                 .child          = binfmt_misc_table,
1489         },
1490 #endif
1491         {
1492                 .procname       = "pipe-max-size",
1493                 .data           = &pipe_max_size,
1494                 .maxlen         = sizeof(int),
1495                 .mode           = 0644,
1496                 .proc_handler   = &pipe_proc_fn,
1497                 .extra1         = &pipe_min_size,
1498         },
1499 /*
1500  * NOTE: do not add new entries to this table unless you have read
1501  * Documentation/sysctl/ctl_unnumbered.txt
1502  */
1503         { }
1504 };
1505
1506 static struct ctl_table debug_table[] = {
1507 #if defined(CONFIG_X86) || defined(CONFIG_PPC) || defined(CONFIG_SPARC) || \
1508     defined(CONFIG_S390)
1509         {
1510                 .procname       = "exception-trace",
1511                 .data           = &show_unhandled_signals,
1512                 .maxlen         = sizeof(int),
1513                 .mode           = 0644,
1514                 .proc_handler   = proc_dointvec
1515         },
1516 #endif
1517 #if defined(CONFIG_OPTPROBES)
1518         {
1519                 .procname       = "kprobes-optimization",
1520                 .data           = &sysctl_kprobes_optimization,
1521                 .maxlen         = sizeof(int),
1522                 .mode           = 0644,
1523                 .proc_handler   = proc_kprobes_optimization_handler,
1524                 .extra1         = &zero,
1525                 .extra2         = &one,
1526         },
1527 #endif
1528         { }
1529 };
1530
1531 static struct ctl_table dev_table[] = {
1532         { }
1533 };
1534
1535 static DEFINE_SPINLOCK(sysctl_lock);
1536
1537 /* called under sysctl_lock */
1538 static int use_table(struct ctl_table_header *p)
1539 {
1540         if (unlikely(p->unregistering))
1541                 return 0;
1542         p->used++;
1543         return 1;
1544 }
1545
1546 /* called under sysctl_lock */
1547 static void unuse_table(struct ctl_table_header *p)
1548 {
1549         if (!--p->used)
1550                 if (unlikely(p->unregistering))
1551                         complete(p->unregistering);
1552 }
1553
1554 /* called under sysctl_lock, will reacquire if has to wait */
1555 static void start_unregistering(struct ctl_table_header *p)
1556 {
1557         /*
1558          * if p->used is 0, nobody will ever touch that entry again;
1559          * we'll eliminate all paths to it before dropping sysctl_lock
1560          */
1561         if (unlikely(p->used)) {
1562                 struct completion wait;
1563                 init_completion(&wait);
1564                 p->unregistering = &wait;
1565                 spin_unlock(&sysctl_lock);
1566                 wait_for_completion(&wait);
1567                 spin_lock(&sysctl_lock);
1568         } else {
1569                 /* anything non-NULL; we'll never dereference it */
1570                 p->unregistering = ERR_PTR(-EINVAL);
1571         }
1572         /*
1573          * do not remove from the list until nobody holds it; walking the
1574          * list in do_sysctl() relies on that.
1575          */
1576         list_del_init(&p->ctl_entry);
1577 }
1578
1579 void sysctl_head_get(struct ctl_table_header *head)
1580 {
1581         spin_lock(&sysctl_lock);
1582         head->count++;
1583         spin_unlock(&sysctl_lock);
1584 }
1585
1586 void sysctl_head_put(struct ctl_table_header *head)
1587 {
1588         spin_lock(&sysctl_lock);
1589         if (!--head->count)
1590                 kfree(head);
1591         spin_unlock(&sysctl_lock);
1592 }
1593
1594 struct ctl_table_header *sysctl_head_grab(struct ctl_table_header *head)
1595 {
1596         if (!head)
1597                 BUG();
1598         spin_lock(&sysctl_lock);
1599         if (!use_table(head))
1600                 head = ERR_PTR(-ENOENT);
1601         spin_unlock(&sysctl_lock);
1602         return head;
1603 }
1604
1605 void sysctl_head_finish(struct ctl_table_header *head)
1606 {
1607         if (!head)
1608                 return;
1609         spin_lock(&sysctl_lock);
1610         unuse_table(head);
1611         spin_unlock(&sysctl_lock);
1612 }
1613
1614 static struct ctl_table_set *
1615 lookup_header_set(struct ctl_table_root *root, struct nsproxy *namespaces)
1616 {
1617         struct ctl_table_set *set = &root->default_set;
1618         if (root->lookup)
1619                 set = root->lookup(root, namespaces);
1620         return set;
1621 }
1622
1623 static struct list_head *
1624 lookup_header_list(struct ctl_table_root *root, struct nsproxy *namespaces)
1625 {
1626         struct ctl_table_set *set = lookup_header_set(root, namespaces);
1627         return &set->list;
1628 }
1629
1630 struct ctl_table_header *__sysctl_head_next(struct nsproxy *namespaces,
1631                                             struct ctl_table_header *prev)
1632 {
1633         struct ctl_table_root *root;
1634         struct list_head *header_list;
1635         struct ctl_table_header *head;
1636         struct list_head *tmp;
1637
1638         spin_lock(&sysctl_lock);
1639         if (prev) {
1640                 head = prev;
1641                 tmp = &prev->ctl_entry;
1642                 unuse_table(prev);
1643                 goto next;
1644         }
1645         tmp = &root_table_header.ctl_entry;
1646         for (;;) {
1647                 head = list_entry(tmp, struct ctl_table_header, ctl_entry);
1648
1649                 if (!use_table(head))
1650                         goto next;
1651                 spin_unlock(&sysctl_lock);
1652                 return head;
1653         next:
1654                 root = head->root;
1655                 tmp = tmp->next;
1656                 header_list = lookup_header_list(root, namespaces);
1657                 if (tmp != header_list)
1658                         continue;
1659
1660                 do {
1661                         root = list_entry(root->root_list.next,
1662                                         struct ctl_table_root, root_list);
1663                         if (root == &sysctl_table_root)
1664                                 goto out;
1665                         header_list = lookup_header_list(root, namespaces);
1666                 } while (list_empty(header_list));
1667                 tmp = header_list->next;
1668         }
1669 out:
1670         spin_unlock(&sysctl_lock);
1671         return NULL;
1672 }
1673
1674 struct ctl_table_header *sysctl_head_next(struct ctl_table_header *prev)
1675 {
1676         return __sysctl_head_next(current->nsproxy, prev);
1677 }
1678
1679 void register_sysctl_root(struct ctl_table_root *root)
1680 {
1681         spin_lock(&sysctl_lock);
1682         list_add_tail(&root->root_list, &sysctl_table_root.root_list);
1683         spin_unlock(&sysctl_lock);
1684 }
1685
1686 /*
1687  * sysctl_perm does NOT grant the superuser all rights automatically, because
1688  * some sysctl variables are readonly even to root.
1689  */
1690
1691 static int test_perm(int mode, int op)
1692 {
1693         if (!current_euid())
1694                 mode >>= 6;
1695         else if (in_egroup_p(0))
1696                 mode >>= 3;
1697         if ((op & ~mode & (MAY_READ|MAY_WRITE|MAY_EXEC)) == 0)
1698                 return 0;
1699         return -EACCES;
1700 }
1701
1702 int sysctl_perm(struct ctl_table_root *root, struct ctl_table *table, int op)
1703 {
1704         int error;
1705         int mode;
1706
1707         error = security_sysctl(table, op & (MAY_READ | MAY_WRITE | MAY_EXEC));
1708         if (error)
1709                 return error;
1710
1711         if (root->permissions)
1712                 mode = root->permissions(root, current->nsproxy, table);
1713         else
1714                 mode = table->mode;
1715
1716         return test_perm(mode, op);
1717 }
1718
1719 static void sysctl_set_parent(struct ctl_table *parent, struct ctl_table *table)
1720 {
1721         for (; table->procname; table++) {
1722                 table->parent = parent;
1723                 if (table->child)
1724                         sysctl_set_parent(table, table->child);
1725         }
1726 }
1727
1728 static __init int sysctl_init(void)
1729 {
1730         sysctl_set_parent(NULL, root_table);
1731 #ifdef CONFIG_SYSCTL_SYSCALL_CHECK
1732         sysctl_check_table(current->nsproxy, root_table);
1733 #endif
1734         return 0;
1735 }
1736
1737 core_initcall(sysctl_init);
1738
1739 static struct ctl_table *is_branch_in(struct ctl_table *branch,
1740                                       struct ctl_table *table)
1741 {
1742         struct ctl_table *p;
1743         const char *s = branch->procname;
1744
1745         /* branch should have named subdirectory as its first element */
1746         if (!s || !branch->child)
1747                 return NULL;
1748
1749         /* ... and nothing else */
1750         if (branch[1].procname)
1751                 return NULL;
1752
1753         /* table should contain subdirectory with the same name */
1754         for (p = table; p->procname; p++) {
1755                 if (!p->child)
1756                         continue;
1757                 if (p->procname && strcmp(p->procname, s) == 0)
1758                         return p;
1759         }
1760         return NULL;
1761 }
1762
1763 /* see if attaching q to p would be an improvement */
1764 static void try_attach(struct ctl_table_header *p, struct ctl_table_header *q)
1765 {
1766         struct ctl_table *to = p->ctl_table, *by = q->ctl_table;
1767         struct ctl_table *next;
1768         int is_better = 0;
1769         int not_in_parent = !p->attached_by;
1770
1771         while ((next = is_branch_in(by, to)) != NULL) {
1772                 if (by == q->attached_by)
1773                         is_better = 1;
1774                 if (to == p->attached_by)
1775                         not_in_parent = 1;
1776                 by = by->child;
1777                 to = next->child;
1778         }
1779
1780         if (is_better && not_in_parent) {
1781                 q->attached_by = by;
1782                 q->attached_to = to;
1783                 q->parent = p;
1784         }
1785 }
1786
1787 /**
1788  * __register_sysctl_paths - register a sysctl hierarchy
1789  * @root: List of sysctl headers to register on
1790  * @namespaces: Data to compute which lists of sysctl entries are visible
1791  * @path: The path to the directory the sysctl table is in.
1792  * @table: the top-level table structure
1793  *
1794  * Register a sysctl table hierarchy. @table should be a filled in ctl_table
1795  * array. A completely 0 filled entry terminates the table.
1796  *
1797  * The members of the &struct ctl_table structure are used as follows:
1798  *
1799  * procname - the name of the sysctl file under /proc/sys. Set to %NULL to not
1800  *            enter a sysctl file
1801  *
1802  * data - a pointer to data for use by proc_handler
1803  *
1804  * maxlen - the maximum size in bytes of the data
1805  *
1806  * mode - the file permissions for the /proc/sys file, and for sysctl(2)
1807  *
1808  * child - a pointer to the child sysctl table if this entry is a directory, or
1809  *         %NULL.
1810  *
1811  * proc_handler - the text handler routine (described below)
1812  *
1813  * de - for internal use by the sysctl routines
1814  *
1815  * extra1, extra2 - extra pointers usable by the proc handler routines
1816  *
1817  * Leaf nodes in the sysctl tree will be represented by a single file
1818  * under /proc; non-leaf nodes will be represented by directories.
1819  *
1820  * sysctl(2) can automatically manage read and write requests through
1821  * the sysctl table.  The data and maxlen fields of the ctl_table
1822  * struct enable minimal validation of the values being written to be
1823  * performed, and the mode field allows minimal authentication.
1824  *
1825  * There must be a proc_handler routine for any terminal nodes
1826  * mirrored under /proc/sys (non-terminals are handled by a built-in
1827  * directory handler).  Several default handlers are available to
1828  * cover common cases -
1829  *
1830  * proc_dostring(), proc_dointvec(), proc_dointvec_jiffies(),
1831  * proc_dointvec_userhz_jiffies(), proc_dointvec_minmax(),
1832  * proc_doulongvec_ms_jiffies_minmax(), proc_doulongvec_minmax()
1833  *
1834  * It is the handler's job to read the input buffer from user memory
1835  * and process it. The handler should return 0 on success.
1836  *
1837  * This routine returns %NULL on a failure to register, and a pointer
1838  * to the table header on success.
1839  */
1840 struct ctl_table_header *__register_sysctl_paths(
1841         struct ctl_table_root *root,
1842         struct nsproxy *namespaces,
1843         const struct ctl_path *path, struct ctl_table *table)
1844 {
1845         struct ctl_table_header *header;
1846         struct ctl_table *new, **prevp;
1847         unsigned int n, npath;
1848         struct ctl_table_set *set;
1849
1850         /* Count the path components */
1851         for (npath = 0; path[npath].procname; ++npath)
1852                 ;
1853
1854         /*
1855          * For each path component, allocate a 2-element ctl_table array.
1856          * The first array element will be filled with the sysctl entry
1857          * for this, the second will be the sentinel (procname == 0).
1858          *
1859          * We allocate everything in one go so that we don't have to
1860          * worry about freeing additional memory in unregister_sysctl_table.
1861          */
1862         header = kzalloc(sizeof(struct ctl_table_header) +
1863                          (2 * npath * sizeof(struct ctl_table)), GFP_KERNEL);
1864         if (!header)
1865                 return NULL;
1866
1867         new = (struct ctl_table *) (header + 1);
1868
1869         /* Now connect the dots */
1870         prevp = &header->ctl_table;
1871         for (n = 0; n < npath; ++n, ++path) {
1872                 /* Copy the procname */
1873                 new->procname = path->procname;
1874                 new->mode     = 0555;
1875
1876                 *prevp = new;
1877                 prevp = &new->child;
1878
1879                 new += 2;
1880         }
1881         *prevp = table;
1882         header->ctl_table_arg = table;
1883
1884         INIT_LIST_HEAD(&header->ctl_entry);
1885         header->used = 0;
1886         header->unregistering = NULL;
1887         header->root = root;
1888         sysctl_set_parent(NULL, header->ctl_table);
1889         header->count = 1;
1890 #ifdef CONFIG_SYSCTL_SYSCALL_CHECK
1891         if (sysctl_check_table(namespaces, header->ctl_table)) {
1892                 kfree(header);
1893                 return NULL;
1894         }
1895 #endif
1896         spin_lock(&sysctl_lock);
1897         header->set = lookup_header_set(root, namespaces);
1898         header->attached_by = header->ctl_table;
1899         header->attached_to = root_table;
1900         header->parent = &root_table_header;
1901         for (set = header->set; set; set = set->parent) {
1902                 struct ctl_table_header *p;
1903                 list_for_each_entry(p, &set->list, ctl_entry) {
1904                         if (p->unregistering)
1905                                 continue;
1906                         try_attach(p, header);
1907                 }
1908         }
1909         header->parent->count++;
1910         list_add_tail(&header->ctl_entry, &header->set->list);
1911         spin_unlock(&sysctl_lock);
1912
1913         return header;
1914 }
1915
1916 /**
1917  * register_sysctl_table_path - register a sysctl table hierarchy
1918  * @path: The path to the directory the sysctl table is in.
1919  * @table: the top-level table structure
1920  *
1921  * Register a sysctl table hierarchy. @table should be a filled in ctl_table
1922  * array. A completely 0 filled entry terminates the table.
1923  *
1924  * See __register_sysctl_paths for more details.
1925  */
1926 struct ctl_table_header *register_sysctl_paths(const struct ctl_path *path,
1927                                                 struct ctl_table *table)
1928 {
1929         return __register_sysctl_paths(&sysctl_table_root, current->nsproxy,
1930                                         path, table);
1931 }
1932
1933 /**
1934  * register_sysctl_table - register a sysctl table hierarchy
1935  * @table: the top-level table structure
1936  *
1937  * Register a sysctl table hierarchy. @table should be a filled in ctl_table
1938  * array. A completely 0 filled entry terminates the table.
1939  *
1940  * See register_sysctl_paths for more details.
1941  */
1942 struct ctl_table_header *register_sysctl_table(struct ctl_table *table)
1943 {
1944         static const struct ctl_path null_path[] = { {} };
1945
1946         return register_sysctl_paths(null_path, table);
1947 }
1948
1949 /**
1950  * unregister_sysctl_table - unregister a sysctl table hierarchy
1951  * @header: the header returned from register_sysctl_table
1952  *
1953  * Unregisters the sysctl table and all children. proc entries may not
1954  * actually be removed until they are no longer used by anyone.
1955  */
1956 void unregister_sysctl_table(struct ctl_table_header * header)
1957 {
1958         might_sleep();
1959
1960         if (header == NULL)
1961                 return;
1962
1963         spin_lock(&sysctl_lock);
1964         start_unregistering(header);
1965         if (!--header->parent->count) {
1966                 WARN_ON(1);
1967                 kfree(header->parent);
1968         }
1969         if (!--header->count)
1970                 kfree(header);
1971         spin_unlock(&sysctl_lock);
1972 }
1973
1974 int sysctl_is_seen(struct ctl_table_header *p)
1975 {
1976         struct ctl_table_set *set = p->set;
1977         int res;
1978         spin_lock(&sysctl_lock);
1979         if (p->unregistering)
1980                 res = 0;
1981         else if (!set->is_seen)
1982                 res = 1;
1983         else
1984                 res = set->is_seen(set);
1985         spin_unlock(&sysctl_lock);
1986         return res;
1987 }
1988
1989 void setup_sysctl_set(struct ctl_table_set *p,
1990         struct ctl_table_set *parent,
1991         int (*is_seen)(struct ctl_table_set *))
1992 {
1993         INIT_LIST_HEAD(&p->list);
1994         p->parent = parent ? parent : &sysctl_table_root.default_set;
1995         p->is_seen = is_seen;
1996 }
1997
1998 #else /* !CONFIG_SYSCTL */
1999 struct ctl_table_header *register_sysctl_table(struct ctl_table * table)
2000 {
2001         return NULL;
2002 }
2003
2004 struct ctl_table_header *register_sysctl_paths(const struct ctl_path *path,
2005                                                     struct ctl_table *table)
2006 {
2007         return NULL;
2008 }
2009
2010 void unregister_sysctl_table(struct ctl_table_header * table)
2011 {
2012 }
2013
2014 void setup_sysctl_set(struct ctl_table_set *p,
2015         struct ctl_table_set *parent,
2016         int (*is_seen)(struct ctl_table_set *))
2017 {
2018 }
2019
2020 void sysctl_head_put(struct ctl_table_header *head)
2021 {
2022 }
2023
2024 #endif /* CONFIG_SYSCTL */
2025
2026 /*
2027  * /proc/sys support
2028  */
2029
2030 #ifdef CONFIG_PROC_SYSCTL
2031
2032 static int _proc_do_string(void* data, int maxlen, int write,
2033                            void __user *buffer,
2034                            size_t *lenp, loff_t *ppos)
2035 {
2036         size_t len;
2037         char __user *p;
2038         char c;
2039
2040         if (!data || !maxlen || !*lenp) {
2041                 *lenp = 0;
2042                 return 0;
2043         }
2044
2045         if (write) {
2046                 len = 0;
2047                 p = buffer;
2048                 while (len < *lenp) {
2049                         if (get_user(c, p++))
2050                                 return -EFAULT;
2051                         if (c == 0 || c == '\n')
2052                                 break;
2053                         len++;
2054                 }
2055                 if (len >= maxlen)
2056                         len = maxlen-1;
2057                 if(copy_from_user(data, buffer, len))
2058                         return -EFAULT;
2059                 ((char *) data)[len] = 0;
2060                 *ppos += *lenp;
2061         } else {
2062                 len = strlen(data);
2063                 if (len > maxlen)
2064                         len = maxlen;
2065
2066                 if (*ppos > len) {
2067                         *lenp = 0;
2068                         return 0;
2069                 }
2070
2071                 data += *ppos;
2072                 len  -= *ppos;
2073
2074                 if (len > *lenp)
2075                         len = *lenp;
2076                 if (len)
2077                         if(copy_to_user(buffer, data, len))
2078                                 return -EFAULT;
2079                 if (len < *lenp) {
2080                         if(put_user('\n', ((char __user *) buffer) + len))
2081                                 return -EFAULT;
2082                         len++;
2083                 }
2084                 *lenp = len;
2085                 *ppos += len;
2086         }
2087         return 0;
2088 }
2089
2090 /**
2091  * proc_dostring - read a string sysctl
2092  * @table: the sysctl table
2093  * @write: %TRUE if this is a write to the sysctl file
2094  * @buffer: the user buffer
2095  * @lenp: the size of the user buffer
2096  * @ppos: file position
2097  *
2098  * Reads/writes a string from/to the user buffer. If the kernel
2099  * buffer provided is not large enough to hold the string, the
2100  * string is truncated. The copied string is %NULL-terminated.
2101  * If the string is being read by the user process, it is copied
2102  * and a newline '\n' is added. It is truncated if the buffer is
2103  * not large enough.
2104  *
2105  * Returns 0 on success.
2106  */
2107 int proc_dostring(struct ctl_table *table, int write,
2108                   void __user *buffer, size_t *lenp, loff_t *ppos)
2109 {
2110         return _proc_do_string(table->data, table->maxlen, write,
2111                                buffer, lenp, ppos);
2112 }
2113
2114 static size_t proc_skip_spaces(char **buf)
2115 {
2116         size_t ret;
2117         char *tmp = skip_spaces(*buf);
2118         ret = tmp - *buf;
2119         *buf = tmp;
2120         return ret;
2121 }
2122
2123 static void proc_skip_char(char **buf, size_t *size, const char v)
2124 {
2125         while (*size) {
2126                 if (**buf != v)
2127                         break;
2128                 (*size)--;
2129                 (*buf)++;
2130         }
2131 }
2132
2133 #define TMPBUFLEN 22
2134 /**
2135  * proc_get_long - reads an ASCII formatted integer from a user buffer
2136  *
2137  * @buf: a kernel buffer
2138  * @size: size of the kernel buffer
2139  * @val: this is where the number will be stored
2140  * @neg: set to %TRUE if number is negative
2141  * @perm_tr: a vector which contains the allowed trailers
2142  * @perm_tr_len: size of the perm_tr vector
2143  * @tr: pointer to store the trailer character
2144  *
2145  * In case of success %0 is returned and @buf and @size are updated with
2146  * the amount of bytes read. If @tr is non-NULL and a trailing
2147  * character exists (size is non-zero after returning from this
2148  * function), @tr is updated with the trailing character.
2149  */
2150 static int proc_get_long(char **buf, size_t *size,
2151                           unsigned long *val, bool *neg,
2152                           const char *perm_tr, unsigned perm_tr_len, char *tr)
2153 {
2154         int len;
2155         char *p, tmp[TMPBUFLEN];
2156
2157         if (!*size)
2158                 return -EINVAL;
2159
2160         len = *size;
2161         if (len > TMPBUFLEN - 1)
2162                 len = TMPBUFLEN - 1;
2163
2164         memcpy(tmp, *buf, len);
2165
2166         tmp[len] = 0;
2167         p = tmp;
2168         if (*p == '-' && *size > 1) {
2169                 *neg = true;
2170                 p++;
2171         } else
2172                 *neg = false;
2173         if (!isdigit(*p))
2174                 return -EINVAL;
2175
2176         *val = simple_strtoul(p, &p, 0);
2177
2178         len = p - tmp;
2179
2180         /* We don't know if the next char is whitespace thus we may accept
2181          * invalid integers (e.g. 1234...a) or two integers instead of one
2182          * (e.g. 123...1). So lets not allow such large numbers. */
2183         if (len == TMPBUFLEN - 1)
2184                 return -EINVAL;
2185
2186         if (len < *size && perm_tr_len && !memchr(perm_tr, *p, perm_tr_len))
2187                 return -EINVAL;
2188
2189         if (tr && (len < *size))
2190                 *tr = *p;
2191
2192         *buf += len;
2193         *size -= len;
2194
2195         return 0;
2196 }
2197
2198 /**
2199  * proc_put_long - converts an integer to a decimal ASCII formatted string
2200  *
2201  * @buf: the user buffer
2202  * @size: the size of the user buffer
2203  * @val: the integer to be converted
2204  * @neg: sign of the number, %TRUE for negative
2205  *
2206  * In case of success %0 is returned and @buf and @size are updated with
2207  * the amount of bytes written.
2208  */
2209 static int proc_put_long(void __user **buf, size_t *size, unsigned long val,
2210                           bool neg)
2211 {
2212         int len;
2213         char tmp[TMPBUFLEN], *p = tmp;
2214
2215         sprintf(p, "%s%lu", neg ? "-" : "", val);
2216         len = strlen(tmp);
2217         if (len > *size)
2218                 len = *size;
2219         if (copy_to_user(*buf, tmp, len))
2220                 return -EFAULT;
2221         *size -= len;
2222         *buf += len;
2223         return 0;
2224 }
2225 #undef TMPBUFLEN
2226
2227 static int proc_put_char(void __user **buf, size_t *size, char c)
2228 {
2229         if (*size) {
2230                 char __user **buffer = (char __user **)buf;
2231                 if (put_user(c, *buffer))
2232                         return -EFAULT;
2233                 (*size)--, (*buffer)++;
2234                 *buf = *buffer;
2235         }
2236         return 0;
2237 }
2238
2239 static int do_proc_dointvec_conv(bool *negp, unsigned long *lvalp,
2240                                  int *valp,
2241                                  int write, void *data)
2242 {
2243         if (write) {
2244                 *valp = *negp ? -*lvalp : *lvalp;
2245         } else {
2246                 int val = *valp;
2247                 if (val < 0) {
2248                         *negp = true;
2249                         *lvalp = (unsigned long)-val;
2250                 } else {
2251                         *negp = false;
2252                         *lvalp = (unsigned long)val;
2253                 }
2254         }
2255         return 0;
2256 }
2257
2258 static const char proc_wspace_sep[] = { ' ', '\t', '\n' };
2259
2260 static int __do_proc_dointvec(void *tbl_data, struct ctl_table *table,
2261                   int write, void __user *buffer,
2262                   size_t *lenp, loff_t *ppos,
2263                   int (*conv)(bool *negp, unsigned long *lvalp, int *valp,
2264                               int write, void *data),
2265                   void *data)
2266 {
2267         int *i, vleft, first = 1, err = 0;
2268         unsigned long page = 0;
2269         size_t left;
2270         char *kbuf;
2271
2272         if (!tbl_data || !table->maxlen || !*lenp || (*ppos && !write)) {
2273                 *lenp = 0;
2274                 return 0;
2275         }
2276
2277         i = (int *) tbl_data;
2278         vleft = table->maxlen / sizeof(*i);
2279         left = *lenp;
2280
2281         if (!conv)
2282                 conv = do_proc_dointvec_conv;
2283
2284         if (write) {
2285                 if (left > PAGE_SIZE - 1)
2286                         left = PAGE_SIZE - 1;
2287                 page = __get_free_page(GFP_TEMPORARY);
2288                 kbuf = (char *) page;
2289                 if (!kbuf)
2290                         return -ENOMEM;
2291                 if (copy_from_user(kbuf, buffer, left)) {
2292                         err = -EFAULT;
2293                         goto free;
2294                 }
2295                 kbuf[left] = 0;
2296         }
2297
2298         for (; left && vleft--; i++, first=0) {
2299                 unsigned long lval;
2300                 bool neg;
2301
2302                 if (write) {
2303                         left -= proc_skip_spaces(&kbuf);
2304
2305                         if (!left)
2306                                 break;
2307                         err = proc_get_long(&kbuf, &left, &lval, &neg,
2308                                              proc_wspace_sep,
2309                                              sizeof(proc_wspace_sep), NULL);
2310                         if (err)
2311                                 break;
2312                         if (conv(&neg, &lval, i, 1, data)) {
2313                                 err = -EINVAL;
2314                                 break;
2315                         }
2316                 } else {
2317                         if (conv(&neg, &lval, i, 0, data)) {
2318                                 err = -EINVAL;
2319                                 break;
2320                         }
2321                         if (!first)
2322                                 err = proc_put_char(&buffer, &left, '\t');
2323                         if (err)
2324                                 break;
2325                         err = proc_put_long(&buffer, &left, lval, neg);
2326                         if (err)
2327                                 break;
2328                 }
2329         }
2330
2331         if (!write && !first && left && !err)
2332                 err = proc_put_char(&buffer, &left, '\n');
2333         if (write && !err && left)
2334                 left -= proc_skip_spaces(&kbuf);
2335 free:
2336         if (write) {
2337                 free_page(page);
2338                 if (first)
2339                         return err ? : -EINVAL;
2340         }
2341         *lenp -= left;
2342         *ppos += *lenp;
2343         return err;
2344 }
2345
2346 static int do_proc_dointvec(struct ctl_table *table, int write,
2347                   void __user *buffer, size_t *lenp, loff_t *ppos,
2348                   int (*conv)(bool *negp, unsigned long *lvalp, int *valp,
2349                               int write, void *data),
2350                   void *data)
2351 {
2352         return __do_proc_dointvec(table->data, table, write,
2353                         buffer, lenp, ppos, conv, data);
2354 }
2355
2356 /**
2357  * proc_dointvec - read a vector of integers
2358  * @table: the sysctl table
2359  * @write: %TRUE if this is a write to the sysctl file
2360  * @buffer: the user buffer
2361  * @lenp: the size of the user buffer
2362  * @ppos: file position
2363  *
2364  * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2365  * values from/to the user buffer, treated as an ASCII string.
2366  *
2367  * Returns 0 on success.
2368  */
2369 int proc_dointvec(struct ctl_table *table, int write,
2370                      void __user *buffer, size_t *lenp, loff_t *ppos)
2371 {
2372     return do_proc_dointvec(table,write,buffer,lenp,ppos,
2373                             NULL,NULL);
2374 }
2375
2376 /*
2377  * Taint values can only be increased
2378  * This means we can safely use a temporary.
2379  */
2380 static int proc_taint(struct ctl_table *table, int write,
2381                                void __user *buffer, size_t *lenp, loff_t *ppos)
2382 {
2383         struct ctl_table t;
2384         unsigned long tmptaint = get_taint();
2385         int err;
2386
2387         if (write && !capable(CAP_SYS_ADMIN))
2388                 return -EPERM;
2389
2390         t = *table;
2391         t.data = &tmptaint;
2392         err = proc_doulongvec_minmax(&t, write, buffer, lenp, ppos);
2393         if (err < 0)
2394                 return err;
2395
2396         if (write) {
2397                 /*
2398                  * Poor man's atomic or. Not worth adding a primitive
2399                  * to everyone's atomic.h for this
2400                  */
2401                 int i;
2402                 for (i = 0; i < BITS_PER_LONG && tmptaint >> i; i++) {
2403                         if ((tmptaint >> i) & 1)
2404                                 add_taint(i);
2405                 }
2406         }
2407
2408         return err;
2409 }
2410
2411 struct do_proc_dointvec_minmax_conv_param {
2412         int *min;
2413         int *max;
2414 };
2415
2416 static int do_proc_dointvec_minmax_conv(bool *negp, unsigned long *lvalp,
2417                                         int *valp,
2418                                         int write, void *data)
2419 {
2420         struct do_proc_dointvec_minmax_conv_param *param = data;
2421         if (write) {
2422                 int val = *negp ? -*lvalp : *lvalp;
2423                 if ((param->min && *param->min > val) ||
2424                     (param->max && *param->max < val))
2425                         return -EINVAL;
2426                 *valp = val;
2427         } else {
2428                 int val = *valp;
2429                 if (val < 0) {
2430                         *negp = true;
2431                         *lvalp = (unsigned long)-val;
2432                 } else {
2433                         *negp = false;
2434                         *lvalp = (unsigned long)val;
2435                 }
2436         }
2437         return 0;
2438 }
2439
2440 /**
2441  * proc_dointvec_minmax - read a vector of integers with min/max values
2442  * @table: the sysctl table
2443  * @write: %TRUE if this is a write to the sysctl file
2444  * @buffer: the user buffer
2445  * @lenp: the size of the user buffer
2446  * @ppos: file position
2447  *
2448  * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2449  * values from/to the user buffer, treated as an ASCII string.
2450  *
2451  * This routine will ensure the values are within the range specified by
2452  * table->extra1 (min) and table->extra2 (max).
2453  *
2454  * Returns 0 on success.
2455  */
2456 int proc_dointvec_minmax(struct ctl_table *table, int write,
2457                   void __user *buffer, size_t *lenp, loff_t *ppos)
2458 {
2459         struct do_proc_dointvec_minmax_conv_param param = {
2460                 .min = (int *) table->extra1,
2461                 .max = (int *) table->extra2,
2462         };
2463         return do_proc_dointvec(table, write, buffer, lenp, ppos,
2464                                 do_proc_dointvec_minmax_conv, &param);
2465 }
2466
2467 static int __do_proc_doulongvec_minmax(void *data, struct ctl_table *table, int write,
2468                                      void __user *buffer,
2469                                      size_t *lenp, loff_t *ppos,
2470                                      unsigned long convmul,
2471                                      unsigned long convdiv)
2472 {
2473         unsigned long *i, *min, *max;
2474         int vleft, first = 1, err = 0;
2475         unsigned long page = 0;
2476         size_t left;
2477         char *kbuf;
2478
2479         if (!data || !table->maxlen || !*lenp || (*ppos && !write)) {
2480                 *lenp = 0;
2481                 return 0;
2482         }
2483
2484         i = (unsigned long *) data;
2485         min = (unsigned long *) table->extra1;
2486         max = (unsigned long *) table->extra2;
2487         vleft = table->maxlen / sizeof(unsigned long);
2488         left = *lenp;
2489
2490         if (write) {
2491                 if (left > PAGE_SIZE - 1)
2492                         left = PAGE_SIZE - 1;
2493                 page = __get_free_page(GFP_TEMPORARY);
2494                 kbuf = (char *) page;
2495                 if (!kbuf)
2496                         return -ENOMEM;
2497                 if (copy_from_user(kbuf, buffer, left)) {
2498                         err = -EFAULT;
2499                         goto free;
2500                 }
2501                 kbuf[left] = 0;
2502         }
2503
2504         for (; left && vleft--; i++, first = 0) {
2505                 unsigned long val;
2506
2507                 if (write) {
2508                         bool neg;
2509
2510                         left -= proc_skip_spaces(&kbuf);
2511
2512                         err = proc_get_long(&kbuf, &left, &val, &neg,
2513                                              proc_wspace_sep,
2514                                              sizeof(proc_wspace_sep), NULL);
2515                         if (err)
2516                                 break;
2517                         if (neg)
2518                                 continue;
2519                         if ((min && val < *min) || (max && val > *max))
2520                                 continue;
2521                         *i = val;
2522                 } else {
2523                         val = convdiv * (*i) / convmul;
2524                         if (!first)
2525                                 err = proc_put_char(&buffer, &left, '\t');
2526                         err = proc_put_long(&buffer, &left, val, false);
2527                         if (err)
2528                                 break;
2529                 }
2530         }
2531
2532         if (!write && !first && left && !err)
2533                 err = proc_put_char(&buffer, &left, '\n');
2534         if (write && !err)
2535                 left -= proc_skip_spaces(&kbuf);
2536 free:
2537         if (write) {
2538                 free_page(page);
2539                 if (first)
2540                         return err ? : -EINVAL;
2541         }
2542         *lenp -= left;
2543         *ppos += *lenp;
2544         return err;
2545 }
2546
2547 static int do_proc_doulongvec_minmax(struct ctl_table *table, int write,
2548                                      void __user *buffer,
2549                                      size_t *lenp, loff_t *ppos,
2550                                      unsigned long convmul,
2551                                      unsigned long convdiv)
2552 {
2553         return __do_proc_doulongvec_minmax(table->data, table, write,
2554                         buffer, lenp, ppos, convmul, convdiv);
2555 }
2556
2557 /**
2558  * proc_doulongvec_minmax - read a vector of long integers with min/max values
2559  * @table: the sysctl table
2560  * @write: %TRUE if this is a write to the sysctl file
2561  * @buffer: the user buffer
2562  * @lenp: the size of the user buffer
2563  * @ppos: file position
2564  *
2565  * Reads/writes up to table->maxlen/sizeof(unsigned long) unsigned long
2566  * values from/to the user buffer, treated as an ASCII string.
2567  *
2568  * This routine will ensure the values are within the range specified by
2569  * table->extra1 (min) and table->extra2 (max).
2570  *
2571  * Returns 0 on success.
2572  */
2573 int proc_doulongvec_minmax(struct ctl_table *table, int write,
2574                            void __user *buffer, size_t *lenp, loff_t *ppos)
2575 {
2576     return do_proc_doulongvec_minmax(table, write, buffer, lenp, ppos, 1l, 1l);
2577 }
2578
2579 /**
2580  * proc_doulongvec_ms_jiffies_minmax - read a vector of millisecond values with min/max values
2581  * @table: the sysctl table
2582  * @write: %TRUE if this is a write to the sysctl file
2583  * @buffer: the user buffer
2584  * @lenp: the size of the user buffer
2585  * @ppos: file position
2586  *
2587  * Reads/writes up to table->maxlen/sizeof(unsigned long) unsigned long
2588  * values from/to the user buffer, treated as an ASCII string. The values
2589  * are treated as milliseconds, and converted to jiffies when they are stored.
2590  *
2591  * This routine will ensure the values are within the range specified by
2592  * table->extra1 (min) and table->extra2 (max).
2593  *
2594  * Returns 0 on success.
2595  */
2596 int proc_doulongvec_ms_jiffies_minmax(struct ctl_table *table, int write,
2597                                       void __user *buffer,
2598                                       size_t *lenp, loff_t *ppos)
2599 {
2600     return do_proc_doulongvec_minmax(table, write, buffer,
2601                                      lenp, ppos, HZ, 1000l);
2602 }
2603
2604
2605 static int do_proc_dointvec_jiffies_conv(bool *negp, unsigned long *lvalp,
2606                                          int *valp,
2607                                          int write, void *data)
2608 {
2609         if (write) {
2610                 if (*lvalp > LONG_MAX / HZ)
2611                         return 1;
2612                 *valp = *negp ? -(*lvalp*HZ) : (*lvalp*HZ);
2613         } else {
2614                 int val = *valp;
2615                 unsigned long lval;
2616                 if (val < 0) {
2617                         *negp = true;
2618                         lval = (unsigned long)-val;
2619                 } else {
2620                         *negp = false;
2621                         lval = (unsigned long)val;
2622                 }
2623                 *lvalp = lval / HZ;
2624         }
2625         return 0;
2626 }
2627
2628 static int do_proc_dointvec_userhz_jiffies_conv(bool *negp, unsigned long *lvalp,
2629                                                 int *valp,
2630                                                 int write, void *data)
2631 {
2632         if (write) {
2633                 if (USER_HZ < HZ && *lvalp > (LONG_MAX / HZ) * USER_HZ)
2634                         return 1;
2635                 *valp = clock_t_to_jiffies(*negp ? -*lvalp : *lvalp);
2636         } else {
2637                 int val = *valp;
2638                 unsigned long lval;
2639                 if (val < 0) {
2640                         *negp = true;
2641                         lval = (unsigned long)-val;
2642                 } else {
2643                         *negp = false;
2644                         lval = (unsigned long)val;
2645                 }
2646                 *lvalp = jiffies_to_clock_t(lval);
2647         }
2648         return 0;
2649 }
2650
2651 static int do_proc_dointvec_ms_jiffies_conv(bool *negp, unsigned long *lvalp,
2652                                             int *valp,
2653                                             int write, void *data)
2654 {
2655         if (write) {
2656                 *valp = msecs_to_jiffies(*negp ? -*lvalp : *lvalp);
2657         } else {
2658                 int val = *valp;
2659                 unsigned long lval;
2660                 if (val < 0) {
2661                         *negp = true;
2662                         lval = (unsigned long)-val;
2663                 } else {
2664                         *negp = false;
2665                         lval = (unsigned long)val;
2666                 }
2667                 *lvalp = jiffies_to_msecs(lval);
2668         }
2669         return 0;
2670 }
2671
2672 /**
2673  * proc_dointvec_jiffies - read a vector of integers as seconds
2674  * @table: the sysctl table
2675  * @write: %TRUE if this is a write to the sysctl file
2676  * @buffer: the user buffer
2677  * @lenp: the size of the user buffer
2678  * @ppos: file position
2679  *
2680  * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2681  * values from/to the user buffer, treated as an ASCII string.
2682  * The values read are assumed to be in seconds, and are converted into
2683  * jiffies.
2684  *
2685  * Returns 0 on success.
2686  */
2687 int proc_dointvec_jiffies(struct ctl_table *table, int write,
2688                           void __user *buffer, size_t *lenp, loff_t *ppos)
2689 {
2690     return do_proc_dointvec(table,write,buffer,lenp,ppos,
2691                             do_proc_dointvec_jiffies_conv,NULL);
2692 }
2693
2694 /**
2695  * proc_dointvec_userhz_jiffies - read a vector of integers as 1/USER_HZ seconds
2696  * @table: the sysctl table
2697  * @write: %TRUE if this is a write to the sysctl file
2698  * @buffer: the user buffer
2699  * @lenp: the size of the user buffer
2700  * @ppos: pointer to the file position
2701  *
2702  * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2703  * values from/to the user buffer, treated as an ASCII string.
2704  * The values read are assumed to be in 1/USER_HZ seconds, and
2705  * are converted into jiffies.
2706  *
2707  * Returns 0 on success.
2708  */
2709 int proc_dointvec_userhz_jiffies(struct ctl_table *table, int write,
2710                                  void __user *buffer, size_t *lenp, loff_t *ppos)
2711 {
2712     return do_proc_dointvec(table,write,buffer,lenp,ppos,
2713                             do_proc_dointvec_userhz_jiffies_conv,NULL);
2714 }
2715
2716 /**
2717  * proc_dointvec_ms_jiffies - read a vector of integers as 1 milliseconds
2718  * @table: the sysctl table
2719  * @write: %TRUE if this is a write to the sysctl file
2720  * @buffer: the user buffer
2721  * @lenp: the size of the user buffer
2722  * @ppos: file position
2723  * @ppos: the current position in the file
2724  *
2725  * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2726  * values from/to the user buffer, treated as an ASCII string.
2727  * The values read are assumed to be in 1/1000 seconds, and
2728  * are converted into jiffies.
2729  *
2730  * Returns 0 on success.
2731  */
2732 int proc_dointvec_ms_jiffies(struct ctl_table *table, int write,
2733                              void __user *buffer, size_t *lenp, loff_t *ppos)
2734 {
2735         return do_proc_dointvec(table, write, buffer, lenp, ppos,
2736                                 do_proc_dointvec_ms_jiffies_conv, NULL);
2737 }
2738
2739 static int proc_do_cad_pid(struct ctl_table *table, int write,
2740                            void __user *buffer, size_t *lenp, loff_t *ppos)
2741 {
2742         struct pid *new_pid;
2743         pid_t tmp;
2744         int r;
2745
2746         tmp = pid_vnr(cad_pid);
2747
2748         r = __do_proc_dointvec(&tmp, table, write, buffer,
2749                                lenp, ppos, NULL, NULL);
2750         if (r || !write)
2751                 return r;
2752
2753         new_pid = find_get_pid(tmp);
2754         if (!new_pid)
2755                 return -ESRCH;
2756
2757         put_pid(xchg(&cad_pid, new_pid));
2758         return 0;
2759 }
2760
2761 /**
2762  * proc_do_large_bitmap - read/write from/to a large bitmap
2763  * @table: the sysctl table
2764  * @write: %TRUE if this is a write to the sysctl file
2765  * @buffer: the user buffer
2766  * @lenp: the size of the user buffer
2767  * @ppos: file position
2768  *
2769  * The bitmap is stored at table->data and the bitmap length (in bits)
2770  * in table->maxlen.
2771  *
2772  * We use a range comma separated format (e.g. 1,3-4,10-10) so that
2773  * large bitmaps may be represented in a compact manner. Writing into
2774  * the file will clear the bitmap then update it with the given input.
2775  *
2776  * Returns 0 on success.
2777  */
2778 int proc_do_large_bitmap(struct ctl_table *table, int write,
2779                          void __user *buffer, size_t *lenp, loff_t *ppos)
2780 {
2781         int err = 0;
2782         bool first = 1;
2783         size_t left = *lenp;
2784         unsigned long bitmap_len = table->maxlen;
2785         unsigned long *bitmap = (unsigned long *) table->data;
2786         unsigned long *tmp_bitmap = NULL;
2787         char tr_a[] = { '-', ',', '\n' }, tr_b[] = { ',', '\n', 0 }, c;
2788
2789         if (!bitmap_len || !left || (*ppos && !write)) {
2790                 *lenp = 0;
2791                 return 0;
2792         }
2793
2794         if (write) {
2795                 unsigned long page = 0;
2796                 char *kbuf;
2797
2798                 if (left > PAGE_SIZE - 1)
2799                         left = PAGE_SIZE - 1;
2800
2801                 page = __get_free_page(GFP_TEMPORARY);
2802                 kbuf = (char *) page;
2803                 if (!kbuf)
2804                         return -ENOMEM;
2805                 if (copy_from_user(kbuf, buffer, left)) {
2806                         free_page(page);
2807                         return -EFAULT;
2808                 }
2809                 kbuf[left] = 0;
2810
2811                 tmp_bitmap = kzalloc(BITS_TO_LONGS(bitmap_len) * sizeof(unsigned long),
2812                                      GFP_KERNEL);
2813                 if (!tmp_bitmap) {
2814                         free_page(page);
2815                         return -ENOMEM;
2816                 }
2817                 proc_skip_char(&kbuf, &left, '\n');
2818                 while (!err && left) {
2819                         unsigned long val_a, val_b;
2820                         bool neg;
2821
2822                         err = proc_get_long(&kbuf, &left, &val_a, &neg, tr_a,
2823                                              sizeof(tr_a), &c);
2824                         if (err)
2825                                 break;
2826                         if (val_a >= bitmap_len || neg) {
2827                                 err = -EINVAL;
2828                                 break;
2829                         }
2830
2831                         val_b = val_a;
2832                         if (left) {
2833                                 kbuf++;
2834                                 left--;
2835                         }
2836
2837                         if (c == '-') {
2838                                 err = proc_get_long(&kbuf, &left, &val_b,
2839                                                      &neg, tr_b, sizeof(tr_b),
2840                                                      &c);
2841                                 if (err)
2842                                         break;
2843                                 if (val_b >= bitmap_len || neg ||
2844                                     val_a > val_b) {
2845                                         err = -EINVAL;
2846                                         break;
2847                                 }
2848                                 if (left) {
2849                                         kbuf++;
2850                                         left--;
2851                                 }
2852                         }
2853
2854                         while (val_a <= val_b)
2855                                 set_bit(val_a++, tmp_bitmap);
2856
2857                         first = 0;
2858                         proc_skip_char(&kbuf, &left, '\n');
2859                 }
2860                 free_page(page);
2861         } else {
2862                 unsigned long bit_a, bit_b = 0;
2863
2864                 while (left) {
2865                         bit_a = find_next_bit(bitmap, bitmap_len, bit_b);
2866                         if (bit_a >= bitmap_len)
2867                                 break;
2868                         bit_b = find_next_zero_bit(bitmap, bitmap_len,
2869                                                    bit_a + 1) - 1;
2870
2871                         if (!first) {
2872                                 err = proc_put_char(&buffer, &left, ',');
2873                                 if (err)
2874                                         break;
2875                         }
2876                         err = proc_put_long(&buffer, &left, bit_a, false);
2877                         if (err)
2878                                 break;
2879                         if (bit_a != bit_b) {
2880                                 err = proc_put_char(&buffer, &left, '-');
2881                                 if (err)
2882                                         break;
2883                                 err = proc_put_long(&buffer, &left, bit_b, false);
2884                                 if (err)
2885                                         break;
2886                         }
2887
2888                         first = 0; bit_b++;
2889                 }
2890                 if (!err)
2891                         err = proc_put_char(&buffer, &left, '\n');
2892         }
2893
2894         if (!err) {
2895                 if (write) {
2896                         if (*ppos)
2897                                 bitmap_or(bitmap, bitmap, tmp_bitmap, bitmap_len);
2898                         else
2899                                 memcpy(bitmap, tmp_bitmap,
2900                                         BITS_TO_LONGS(bitmap_len) * sizeof(unsigned long));
2901                 }
2902                 kfree(tmp_bitmap);
2903                 *lenp -= left;
2904                 *ppos += *lenp;
2905                 return 0;
2906         } else {
2907                 kfree(tmp_bitmap);
2908                 return err;
2909         }
2910 }
2911
2912 #else /* CONFIG_PROC_FS */
2913
2914 int proc_dostring(struct ctl_table *table, int write,
2915                   void __user *buffer, size_t *lenp, loff_t *ppos)
2916 {
2917         return -ENOSYS;
2918 }
2919
2920 int proc_dointvec(struct ctl_table *table, int write,
2921                   void __user *buffer, size_t *lenp, loff_t *ppos)
2922 {
2923         return -ENOSYS;
2924 }
2925
2926 int proc_dointvec_minmax(struct ctl_table *table, int write,
2927                     void __user *buffer, size_t *lenp, loff_t *ppos)
2928 {
2929         return -ENOSYS;
2930 }
2931
2932 int proc_dointvec_jiffies(struct ctl_table *table, int write,
2933                     void __user *buffer, size_t *lenp, loff_t *ppos)
2934 {
2935         return -ENOSYS;
2936 }
2937
2938 int proc_dointvec_userhz_jiffies(struct ctl_table *table, int write,
2939                     void __user *buffer, size_t *lenp, loff_t *ppos)
2940 {
2941         return -ENOSYS;
2942 }
2943
2944 int proc_dointvec_ms_jiffies(struct ctl_table *table, int write,
2945                              void __user *buffer, size_t *lenp, loff_t *ppos)
2946 {
2947         return -ENOSYS;
2948 }
2949
2950 int proc_doulongvec_minmax(struct ctl_table *table, int write,
2951                     void __user *buffer, size_t *lenp, loff_t *ppos)
2952 {
2953         return -ENOSYS;
2954 }
2955
2956 int proc_doulongvec_ms_jiffies_minmax(struct ctl_table *table, int write,
2957                                       void __user *buffer,
2958                                       size_t *lenp, loff_t *ppos)
2959 {
2960     return -ENOSYS;
2961 }
2962
2963
2964 #endif /* CONFIG_PROC_FS */
2965
2966 /*
2967  * No sense putting this after each symbol definition, twice,
2968  * exception granted :-)
2969  */
2970 EXPORT_SYMBOL(proc_dointvec);
2971 EXPORT_SYMBOL(proc_dointvec_jiffies);
2972 EXPORT_SYMBOL(proc_dointvec_minmax);
2973 EXPORT_SYMBOL(proc_dointvec_userhz_jiffies);
2974 EXPORT_SYMBOL(proc_dointvec_ms_jiffies);
2975 EXPORT_SYMBOL(proc_dostring);
2976 EXPORT_SYMBOL(proc_doulongvec_minmax);
2977 EXPORT_SYMBOL(proc_doulongvec_ms_jiffies_minmax);
2978 EXPORT_SYMBOL(register_sysctl_table);
2979 EXPORT_SYMBOL(register_sysctl_paths);
2980 EXPORT_SYMBOL(unregister_sysctl_table);