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 int __maybe_unused three = 3;
 121 static unsigned long one_ul = 1;
 122 static int one_hundred = 100;
 123 #ifdef CONFIG_PRINTK
 124 static int ten_thousand = 10000;
 125 #endif
 126
 127 /* this is needed for the proc_doulongvec_minmax of vm_dirty_bytes */
 128 static unsigned long dirty_bytes_min = 2 * PAGE_SIZE;
 129
 130 /* this is needed for the proc_dointvec_minmax for [fs_]overflow UID and GID */
 131 static int maxolduid = 65535;
 132 static int minolduid;
 133 static int min_percpu_pagelist_fract = 8;
 134
 135 static int ngroups_max = NGROUPS_MAX;
 136
 137 #ifdef CONFIG_INOTIFY_USER
 138 #include <linux/inotify.h>
 139 #endif
 140 #ifdef CONFIG_SPARC
 141 #include <asm/system.h>
 142 #endif
 143
 144 #ifdef CONFIG_SPARC64
 145 extern int sysctl_tsb_ratio;
 146 #endif
 147
 148 #ifdef __hppa__
 149 extern int pwrsw_enabled;
 150 extern int unaligned_enabled;
 151 #endif
 152
 153 #ifdef CONFIG_S390
 154 #ifdef CONFIG_MATHEMU
 155 extern int sysctl_ieee_emulation_warnings;
 156 #endif
 157 extern int sysctl_userprocess_debug;
 158 extern int spin_retry;
 159 #endif
 160
 161 #ifdef CONFIG_IA64
 162 extern int no_unaligned_warning;
 163 extern int unaligned_dump_stack;
 164 #endif
 165
 166 #ifdef CONFIG_PROC_SYSCTL
 167 static int proc_do_cad_pid(struct ctl_table *table, int write,
 168                   void __user *buffer, size_t *lenp, loff_t *ppos);
 169 static int proc_taint(struct ctl_table *table, int write,
 170                                void __user *buffer, size_t *lenp, loff_t *ppos);
 171 #endif
 172
 173 #ifdef CONFIG_PRINTK
 174 static int proc_dmesg_restrict(struct ctl_table *table, int write,
 175                                 void __user *buffer, size_t *lenp, loff_t *ppos);
 176 #endif
 177
 178 #ifdef CONFIG_MAGIC_SYSRQ
 179 /* Note: sysrq code uses it's own private copy */
 180 static int __sysrq_enabled = SYSRQ_DEFAULT_ENABLE;
 181
 182 static int sysrq_sysctl_handler(ctl_table *table, int write,
 183                                 void __user *buffer, size_t *lenp,
 184                                 loff_t *ppos)
 185 {
 186         int error;
 187
 188         error = proc_dointvec(table, write, buffer, lenp, ppos);
 189         if (error)
 190                 return error;
 191
 192         if (write)
 193                 sysrq_toggle_support(__sysrq_enabled);
 194
 195         return 0;
 196 }
 197
 198 #endif
 199
 200 static struct ctl_table root_table[];
 201 static struct ctl_table_root sysctl_table_root;
 202 static struct ctl_table_header root_table_header = {
 203         {{.count = 1,
 204         .ctl_table = root_table,
 205         .ctl_entry = LIST_HEAD_INIT(sysctl_table_root.default_set.list),}},
 206         .root = &sysctl_table_root,
 207         .set = &sysctl_table_root.default_set,
 208 };
 209 static struct ctl_table_root sysctl_table_root = {
 210         .root_list = LIST_HEAD_INIT(sysctl_table_root.root_list),
 211         .default_set.list = LIST_HEAD_INIT(root_table_header.ctl_entry),
 212 };
 213
 214 static struct ctl_table kern_table[];
 215 static struct ctl_table vm_table[];
 216 static struct ctl_table fs_table[];
 217 static struct ctl_table debug_table[];
 218 static struct ctl_table dev_table[];
 219 extern struct ctl_table random_table[];
 220 #ifdef CONFIG_EPOLL
 221 extern struct ctl_table epoll_table[];
 222 #endif
 223
 224 #ifdef HAVE_ARCH_PICK_MMAP_LAYOUT
 225 int sysctl_legacy_va_layout;
 226 #endif
 227
 228 /* The default sysctl tables: */
 229
 230 static struct ctl_table root_table[] = {
 231         {
 232                 .procname       = "kernel",
 233                 .mode           = 0555,
 234                 .child          = kern_table,
 235         },
 236         {
 237                 .procname       = "vm",
 238                 .mode           = 0555,
 239                 .child          = vm_table,
 240         },
 241         {
 242                 .procname       = "fs",
 243                 .mode           = 0555,
 244                 .child          = fs_table,
 245         },
 246         {
 247                 .procname       = "debug",
 248                 .mode           = 0555,
 249                 .child          = debug_table,
 250         },
 251         {
 252                 .procname       = "dev",
 253                 .mode           = 0555,
 254                 .child          = dev_table,
 255         },
 256         { }
 257 };
 258
 259 #ifdef CONFIG_SCHED_DEBUG
 260 static int min_sched_granularity_ns = 100000;           /* 100 usecs */
 261 static int max_sched_granularity_ns = NSEC_PER_SEC;     /* 1 second */
 262 static int min_wakeup_granularity_ns;                   /* 0 usecs */
 263 static int max_wakeup_granularity_ns = NSEC_PER_SEC;    /* 1 second */
 264 static int min_sched_tunable_scaling = SCHED_TUNABLESCALING_NONE;
 265 static int max_sched_tunable_scaling = SCHED_TUNABLESCALING_END-1;
 266 #endif
 267
 268 #ifdef CONFIG_COMPACTION
 269 static int min_extfrag_threshold;
 270 static int max_extfrag_threshold = 1000;
 271 #endif
 272
 273 static struct ctl_table kern_table[] = {
 274         {
 275                 .procname       = "sched_child_runs_first",
 276                 .data           = &sysctl_sched_child_runs_first,
 277                 .maxlen         = sizeof(unsigned int),
 278                 .mode           = 0644,
 279                 .proc_handler   = proc_dointvec,
 280         },
 281 #ifdef CONFIG_SCHED_DEBUG
 282         {
 283                 .procname       = "sched_min_granularity_ns",
 284                 .data           = &sysctl_sched_min_granularity,
 285                 .maxlen         = sizeof(unsigned int),
 286                 .mode           = 0644,
 287                 .proc_handler   = sched_proc_update_handler,
 288                 .extra1         = &min_sched_granularity_ns,
 289                 .extra2         = &max_sched_granularity_ns,
 290         },
 291         {
 292                 .procname       = "sched_latency_ns",
 293                 .data           = &sysctl_sched_latency,
 294                 .maxlen         = sizeof(unsigned int),
 295                 .mode           = 0644,
 296                 .proc_handler   = sched_proc_update_handler,
 297                 .extra1         = &min_sched_granularity_ns,
 298                 .extra2         = &max_sched_granularity_ns,
 299         },
 300         {
 301                 .procname       = "sched_wakeup_granularity_ns",
 302                 .data           = &sysctl_sched_wakeup_granularity,
 303                 .maxlen         = sizeof(unsigned int),
 304                 .mode           = 0644,
 305                 .proc_handler   = sched_proc_update_handler,
 306                 .extra1         = &min_wakeup_granularity_ns,
 307                 .extra2         = &max_wakeup_granularity_ns,
 308         },
 309         {
 310                 .procname       = "sched_tunable_scaling",
 311                 .data           = &sysctl_sched_tunable_scaling,
 312                 .maxlen         = sizeof(enum sched_tunable_scaling),
 313                 .mode           = 0644,
 314                 .proc_handler   = sched_proc_update_handler,
 315                 .extra1         = &min_sched_tunable_scaling,
 316                 .extra2         = &max_sched_tunable_scaling,
 317         },
 318         {
 319                 .procname       = "sched_migration_cost",
 320                 .data           = &sysctl_sched_migration_cost,
 321                 .maxlen         = sizeof(unsigned int),
 322                 .mode           = 0644,
 323                 .proc_handler   = proc_dointvec,
 324         },
 325         {
 326                 .procname       = "sched_nr_migrate",
 327                 .data           = &sysctl_sched_nr_migrate,
 328                 .maxlen         = sizeof(unsigned int),
 329                 .mode           = 0644,
 330                 .proc_handler   = proc_dointvec,
 331         },
 332         {
 333                 .procname       = "sched_time_avg",
 334                 .data           = &sysctl_sched_time_avg,
 335                 .maxlen         = sizeof(unsigned int),
 336                 .mode           = 0644,
 337                 .proc_handler   = proc_dointvec,
 338         },
 339         {
 340                 .procname       = "sched_shares_window",
 341                 .data           = &sysctl_sched_shares_window,
 342                 .maxlen         = sizeof(unsigned int),
 343                 .mode           = 0644,
 344                 .proc_handler   = proc_dointvec,
 345         },
 346         {
 347                 .procname       = "timer_migration",
 348                 .data           = &sysctl_timer_migration,
 349                 .maxlen         = sizeof(unsigned int),
 350                 .mode           = 0644,
 351                 .proc_handler   = proc_dointvec_minmax,
 352                 .extra1         = &zero,
 353                 .extra2         = &one,
 354         },
 355 #endif
 356         {
 357                 .procname       = "sched_rt_period_us",
 358                 .data           = &sysctl_sched_rt_period,
 359                 .maxlen         = sizeof(unsigned int),
 360                 .mode           = 0644,
 361                 .proc_handler   = sched_rt_handler,
 362         },
 363         {
 364                 .procname       = "sched_rt_runtime_us",
 365                 .data           = &sysctl_sched_rt_runtime,
 366                 .maxlen         = sizeof(int),
 367                 .mode           = 0644,
 368                 .proc_handler   = sched_rt_handler,
 369         },
 370 #ifdef CONFIG_SCHED_AUTOGROUP
 371         {
 372                 .procname       = "sched_autogroup_enabled",
 373                 .data           = &sysctl_sched_autogroup_enabled,
 374                 .maxlen         = sizeof(unsigned int),
 375                 .mode           = 0644,
 376                 .proc_handler   = proc_dointvec_minmax,
 377                 .extra1         = &zero,
 378                 .extra2         = &one,
 379         },
 380 #endif
 381 #ifdef CONFIG_PROVE_LOCKING
 382         {
 383                 .procname       = "prove_locking",
 384                 .data           = &prove_locking,
 385                 .maxlen         = sizeof(int),
 386                 .mode           = 0644,
 387                 .proc_handler   = proc_dointvec,
 388         },
 389 #endif
 390 #ifdef CONFIG_LOCK_STAT
 391         {
 392                 .procname       = "lock_stat",
 393                 .data           = &lock_stat,
 394                 .maxlen         = sizeof(int),
 395                 .mode           = 0644,
 396                 .proc_handler   = proc_dointvec,
 397         },
 398 #endif
 399         {
 400                 .procname       = "panic",
 401                 .data           = &panic_timeout,
 402                 .maxlen         = sizeof(int),
 403                 .mode           = 0644,
 404                 .proc_handler   = proc_dointvec,
 405         },
 406         {
 407                 .procname       = "core_uses_pid",
 408                 .data           = &core_uses_pid,
 409                 .maxlen         = sizeof(int),
 410                 .mode           = 0644,
 411                 .proc_handler   = proc_dointvec,
 412         },
 413         {
 414                 .procname       = "core_pattern",
 415                 .data           = core_pattern,
 416                 .maxlen         = CORENAME_MAX_SIZE,
 417                 .mode           = 0644,
 418                 .proc_handler   = proc_dostring,
 419         },
 420         {
 421                 .procname       = "core_pipe_limit",
 422                 .data           = &core_pipe_limit,
 423                 .maxlen         = sizeof(unsigned int),
 424                 .mode           = 0644,
 425                 .proc_handler   = proc_dointvec,
 426         },
 427 #ifdef CONFIG_PROC_SYSCTL
 428         {
 429                 .procname       = "tainted",
 430                 .maxlen         = sizeof(long),
 431                 .mode           = 0644,
 432                 .proc_handler   = proc_taint,
 433         },
 434 #endif
 435 #ifdef CONFIG_LATENCYTOP
 436         {
 437                 .procname       = "latencytop",
 438                 .data           = &latencytop_enabled,
 439                 .maxlen         = sizeof(int),
 440                 .mode           = 0644,
 441                 .proc_handler   = proc_dointvec,
 442         },
 443 #endif
 444 #ifdef CONFIG_BLK_DEV_INITRD
 445         {
 446                 .procname       = "real-root-dev",
 447                 .data           = &real_root_dev,
 448                 .maxlen         = sizeof(int),
 449                 .mode           = 0644,
 450                 .proc_handler   = proc_dointvec,
 451         },
 452 #endif
 453         {
 454                 .procname       = "print-fatal-signals",
 455                 .data           = &print_fatal_signals,
 456                 .maxlen         = sizeof(int),
 457                 .mode           = 0644,
 458                 .proc_handler   = proc_dointvec,
 459         },
 460 #ifdef CONFIG_SPARC
 461         {
 462                 .procname       = "reboot-cmd",
 463                 .data           = reboot_command,
 464                 .maxlen         = 256,
 465                 .mode           = 0644,
 466                 .proc_handler   = proc_dostring,
 467         },
 468         {
 469                 .procname       = "stop-a",
 470                 .data           = &stop_a_enabled,
 471                 .maxlen         = sizeof (int),
 472                 .mode           = 0644,
 473                 .proc_handler   = proc_dointvec,
 474         },
 475         {
 476                 .procname       = "scons-poweroff",
 477                 .data           = &scons_pwroff,
 478                 .maxlen         = sizeof (int),
 479                 .mode           = 0644,
 480                 .proc_handler   = proc_dointvec,
 481         },
 482 #endif
 483 #ifdef CONFIG_SPARC64
 484         {
 485                 .procname       = "tsb-ratio",
 486                 .data           = &sysctl_tsb_ratio,
 487                 .maxlen         = sizeof (int),
 488                 .mode           = 0644,
 489                 .proc_handler   = proc_dointvec,
 490         },
 491 #endif
 492 #ifdef __hppa__
 493         {
 494                 .procname       = "soft-power",
 495                 .data           = &pwrsw_enabled,
 496                 .maxlen         = sizeof (int),
 497                 .mode           = 0644,
 498                 .proc_handler   = proc_dointvec,
 499         },
 500         {
 501                 .procname       = "unaligned-trap",
 502                 .data           = &unaligned_enabled,
 503                 .maxlen         = sizeof (int),
 504                 .mode           = 0644,
 505                 .proc_handler   = proc_dointvec,
 506         },
 507 #endif
 508         {
 509                 .procname       = "ctrl-alt-del",
 510                 .data           = &C_A_D,
 511                 .maxlen         = sizeof(int),
 512                 .mode           = 0644,
 513                 .proc_handler   = proc_dointvec,
 514         },
 515 #ifdef CONFIG_FUNCTION_TRACER
 516         {
 517                 .procname       = "ftrace_enabled",
 518                 .data           = &ftrace_enabled,
 519                 .maxlen         = sizeof(int),
 520                 .mode           = 0644,
 521                 .proc_handler   = ftrace_enable_sysctl,
 522         },
 523 #endif
 524 #ifdef CONFIG_STACK_TRACER
 525         {
 526                 .procname       = "stack_tracer_enabled",
 527                 .data           = &stack_tracer_enabled,
 528                 .maxlen         = sizeof(int),
 529                 .mode           = 0644,
 530                 .proc_handler   = stack_trace_sysctl,
 531         },
 532 #endif
 533 #ifdef CONFIG_TRACING
 534         {
 535                 .procname       = "ftrace_dump_on_oops",
 536                 .data           = &ftrace_dump_on_oops,
 537                 .maxlen         = sizeof(int),
 538                 .mode           = 0644,
 539                 .proc_handler   = proc_dointvec,
 540         },
 541 #endif
 542 #ifdef CONFIG_MODULES
 543         {
 544                 .procname       = "modprobe",
 545                 .data           = &modprobe_path,
 546                 .maxlen         = KMOD_PATH_LEN,
 547                 .mode           = 0644,
 548                 .proc_handler   = proc_dostring,
 549         },
 550         {
 551                 .procname       = "modules_disabled",
 552                 .data           = &modules_disabled,
 553                 .maxlen         = sizeof(int),
 554                 .mode           = 0644,
 555                 /* only handle a transition from default "0" to "1" */
 556                 .proc_handler   = proc_dointvec_minmax,
 557                 .extra1         = &one,
 558                 .extra2         = &one,
 559         },
 560 #endif
 561 #ifdef CONFIG_HOTPLUG
 562         {
 563                 .procname       = "hotplug",
 564                 .data           = &uevent_helper,
 565                 .maxlen         = UEVENT_HELPER_PATH_LEN,
 566                 .mode           = 0644,
 567                 .proc_handler   = proc_dostring,
 568         },
 569 #endif
 570 #ifdef CONFIG_CHR_DEV_SG
 571         {
 572                 .procname       = "sg-big-buff",
 573                 .data           = &sg_big_buff,
 574                 .maxlen         = sizeof (int),
 575                 .mode           = 0444,
 576                 .proc_handler   = proc_dointvec,
 577         },
 578 #endif
 579 #ifdef CONFIG_BSD_PROCESS_ACCT
 580         {
 581                 .procname       = "acct",
 582                 .data           = &acct_parm,
 583                 .maxlen         = 3*sizeof(int),
 584                 .mode           = 0644,
 585                 .proc_handler   = proc_dointvec,
 586         },
 587 #endif
 588 #ifdef CONFIG_MAGIC_SYSRQ
 589         {
 590                 .procname       = "sysrq",
 591                 .data           = &__sysrq_enabled,
 592                 .maxlen         = sizeof (int),
 593                 .mode           = 0644,
 594                 .proc_handler   = sysrq_sysctl_handler,
 595         },
 596 #endif
 597 #ifdef CONFIG_PROC_SYSCTL
 598         {
 599                 .procname       = "cad_pid",
 600                 .data           = NULL,
 601                 .maxlen         = sizeof (int),
 602                 .mode           = 0600,
 603                 .proc_handler   = proc_do_cad_pid,
 604         },
 605 #endif
 606         {
 607                 .procname       = "threads-max",
 608                 .data           = &max_threads,
 609                 .maxlen         = sizeof(int),
 610                 .mode           = 0644,
 611                 .proc_handler   = proc_dointvec,
 612         },
 613         {
 614                 .procname       = "random",
 615                 .mode           = 0555,
 616                 .child          = random_table,
 617         },
 618         {
 619                 .procname       = "overflowuid",
 620                 .data           = &overflowuid,
 621                 .maxlen         = sizeof(int),
 622                 .mode           = 0644,
 623                 .proc_handler   = proc_dointvec_minmax,
 624                 .extra1         = &minolduid,
 625                 .extra2         = &maxolduid,
 626         },
 627         {
 628                 .procname       = "overflowgid",
 629                 .data           = &overflowgid,
 630                 .maxlen         = sizeof(int),
 631                 .mode           = 0644,
 632                 .proc_handler   = proc_dointvec_minmax,
 633                 .extra1         = &minolduid,
 634                 .extra2         = &maxolduid,
 635         },
 636 #ifdef CONFIG_S390
 637 #ifdef CONFIG_MATHEMU
 638         {
 639                 .procname       = "ieee_emulation_warnings",
 640                 .data           = &sysctl_ieee_emulation_warnings,
 641                 .maxlen         = sizeof(int),
 642                 .mode           = 0644,
 643                 .proc_handler   = proc_dointvec,
 644         },
 645 #endif
 646         {
 647                 .procname       = "userprocess_debug",
 648                 .data           = &show_unhandled_signals,
 649                 .maxlen         = sizeof(int),
 650                 .mode           = 0644,
 651                 .proc_handler   = proc_dointvec,
 652         },
 653 #endif
 654         {
 655                 .procname       = "pid_max",
 656                 .data           = &pid_max,
 657                 .maxlen         = sizeof (int),
 658                 .mode           = 0644,
 659                 .proc_handler   = proc_dointvec_minmax,
 660                 .extra1         = &pid_max_min,
 661                 .extra2         = &pid_max_max,
 662         },
 663         {
 664                 .procname       = "panic_on_oops",
 665                 .data           = &panic_on_oops,
 666                 .maxlen         = sizeof(int),
 667                 .mode           = 0644,
 668                 .proc_handler   = proc_dointvec,
 669         },
 670 #if defined CONFIG_PRINTK
 671         {
 672                 .procname       = "printk",
 673                 .data           = &console_loglevel,
 674                 .maxlen         = 4*sizeof(int),
 675                 .mode           = 0644,
 676                 .proc_handler   = proc_dointvec,
 677         },
 678         {
 679                 .procname       = "printk_ratelimit",
 680                 .data           = &printk_ratelimit_state.interval,
 681                 .maxlen         = sizeof(int),
 682                 .mode           = 0644,
 683                 .proc_handler   = proc_dointvec_jiffies,
 684         },
 685         {
 686                 .procname       = "printk_ratelimit_burst",
 687                 .data           = &printk_ratelimit_state.burst,
 688                 .maxlen         = sizeof(int),
 689                 .mode           = 0644,
 690                 .proc_handler   = proc_dointvec,
 691         },
 692         {
 693                 .procname       = "printk_delay",
 694                 .data           = &printk_delay_msec,
 695                 .maxlen         = sizeof(int),
 696                 .mode           = 0644,
 697                 .proc_handler   = proc_dointvec_minmax,
 698                 .extra1         = &zero,
 699                 .extra2         = &ten_thousand,
 700         },
 701         {
 702                 .procname       = "dmesg_restrict",
 703                 .data           = &dmesg_restrict,
 704                 .maxlen         = sizeof(int),
 705                 .mode           = 0644,
 706                 .proc_handler   = proc_dointvec_minmax,
 707                 .extra1         = &zero,
 708                 .extra2         = &one,
 709         },
 710         {
 711                 .procname       = "kptr_restrict",
 712                 .data           = &kptr_restrict,
 713                 .maxlen         = sizeof(int),
 714                 .mode           = 0644,
 715                 .proc_handler   = proc_dmesg_restrict,
 716                 .extra1         = &zero,
 717                 .extra2         = &two,
 718         },
 719 #endif
 720         {
 721                 .procname       = "ngroups_max",
 722                 .data           = &ngroups_max,
 723                 .maxlen         = sizeof (int),
 724                 .mode           = 0444,
 725                 .proc_handler   = proc_dointvec,
 726         },
 727 #if defined(CONFIG_LOCKUP_DETECTOR)
 728         {
 729                 .procname       = "watchdog",
 730                 .data           = &watchdog_enabled,
 731                 .maxlen         = sizeof (int),
 732                 .mode           = 0644,
 733                 .proc_handler   = proc_dowatchdog_enabled,
 734         },
 735         {
 736                 .procname       = "watchdog_thresh",
 737                 .data           = &softlockup_thresh,
 738                 .maxlen         = sizeof(int),
 739                 .mode           = 0644,
 740                 .proc_handler   = proc_dowatchdog_thresh,
 741                 .extra1         = &neg_one,
 742                 .extra2         = &sixty,
 743         },
 744         {
 745                 .procname       = "softlockup_panic",
 746                 .data           = &softlockup_panic,
 747                 .maxlen         = sizeof(int),
 748                 .mode           = 0644,
 749                 .proc_handler   = proc_dointvec_minmax,
 750                 .extra1         = &zero,
 751                 .extra2         = &one,
 752         },
 753         {
 754                 .procname       = "nmi_watchdog",
 755                 .data           = &watchdog_enabled,
 756                 .maxlen         = sizeof (int),
 757                 .mode           = 0644,
 758                 .proc_handler   = proc_dowatchdog_enabled,
 759         },
 760 #endif
 761 #if defined(CONFIG_X86_LOCAL_APIC) && defined(CONFIG_X86)
 762         {
 763                 .procname       = "unknown_nmi_panic",
 764                 .data           = &unknown_nmi_panic,
 765                 .maxlen         = sizeof (int),
 766                 .mode           = 0644,
 767                 .proc_handler   = proc_dointvec,
 768         },
 769 #endif
 770 #if defined(CONFIG_X86)
 771         {
 772                 .procname       = "panic_on_unrecovered_nmi",
 773                 .data           = &panic_on_unrecovered_nmi,
 774                 .maxlen         = sizeof(int),
 775                 .mode           = 0644,
 776                 .proc_handler   = proc_dointvec,
 777         },
 778         {
 779                 .procname       = "panic_on_io_nmi",
 780                 .data           = &panic_on_io_nmi,
 781                 .maxlen         = sizeof(int),
 782                 .mode           = 0644,
 783                 .proc_handler   = proc_dointvec,
 784         },
 785         {
 786                 .procname       = "bootloader_type",
 787                 .data           = &bootloader_type,
 788                 .maxlen         = sizeof (int),
 789                 .mode           = 0444,
 790                 .proc_handler   = proc_dointvec,
 791         },
 792         {
 793                 .procname       = "bootloader_version",
 794                 .data           = &bootloader_version,
 795                 .maxlen         = sizeof (int),
 796                 .mode           = 0444,
 797                 .proc_handler   = proc_dointvec,
 798         },
 799         {
 800                 .procname       = "kstack_depth_to_print",
 801                 .data           = &kstack_depth_to_print,
 802                 .maxlen         = sizeof(int),
 803                 .mode           = 0644,
 804                 .proc_handler   = proc_dointvec,
 805         },
 806         {
 807                 .procname       = "io_delay_type",
 808                 .data           = &io_delay_type,
 809                 .maxlen         = sizeof(int),
 810                 .mode           = 0644,
 811                 .proc_handler   = proc_dointvec,
 812         },
 813 #endif
 814 #if defined(CONFIG_MMU)
 815         {
 816                 .procname       = "randomize_va_space",
 817                 .data           = &randomize_va_space,
 818                 .maxlen         = sizeof(int),
 819                 .mode           = 0644,
 820                 .proc_handler   = proc_dointvec,
 821         },
 822 #endif
 823 #if defined(CONFIG_S390) && defined(CONFIG_SMP)
 824         {
 825                 .procname       = "spin_retry",
 826                 .data           = &spin_retry,
 827                 .maxlen         = sizeof (int),
 828                 .mode           = 0644,
 829                 .proc_handler   = proc_dointvec,
 830         },
 831 #endif
 832 #if     defined(CONFIG_ACPI_SLEEP) && defined(CONFIG_X86)
 833         {
 834                 .procname       = "acpi_video_flags",
 835                 .data           = &acpi_realmode_flags,
 836                 .maxlen         = sizeof (unsigned long),
 837                 .mode           = 0644,
 838                 .proc_handler   = proc_doulongvec_minmax,
 839         },
 840 #endif
 841 #ifdef CONFIG_IA64
 842         {
 843                 .procname       = "ignore-unaligned-usertrap",
 844                 .data           = &no_unaligned_warning,
 845                 .maxlen         = sizeof (int),
 846                 .mode           = 0644,
 847                 .proc_handler   = proc_dointvec,
 848         },
 849         {
 850                 .procname       = "unaligned-dump-stack",
 851                 .data           = &unaligned_dump_stack,
 852                 .maxlen         = sizeof (int),
 853                 .mode           = 0644,
 854                 .proc_handler   = proc_dointvec,
 855         },
 856 #endif
 857 #ifdef CONFIG_DETECT_HUNG_TASK
 858         {
 859                 .procname       = "hung_task_panic",
 860                 .data           = &sysctl_hung_task_panic,
 861                 .maxlen         = sizeof(int),
 862                 .mode           = 0644,
 863                 .proc_handler   = proc_dointvec_minmax,
 864                 .extra1         = &zero,
 865                 .extra2         = &one,
 866         },
 867         {
 868                 .procname       = "hung_task_check_count",
 869                 .data           = &sysctl_hung_task_check_count,
 870                 .maxlen         = sizeof(unsigned long),
 871                 .mode           = 0644,
 872                 .proc_handler   = proc_doulongvec_minmax,
 873         },
 874         {
 875                 .procname       = "hung_task_timeout_secs",
 876                 .data           = &sysctl_hung_task_timeout_secs,
 877                 .maxlen         = sizeof(unsigned long),
 878                 .mode           = 0644,
 879                 .proc_handler   = proc_dohung_task_timeout_secs,
 880         },
 881         {
 882                 .procname       = "hung_task_warnings",
 883                 .data           = &sysctl_hung_task_warnings,
 884                 .maxlen         = sizeof(unsigned long),
 885                 .mode           = 0644,
 886                 .proc_handler   = proc_doulongvec_minmax,
 887         },
 888 #endif
 889 #ifdef CONFIG_COMPAT
 890         {
 891                 .procname       = "compat-log",
 892                 .data           = &compat_log,
 893                 .maxlen         = sizeof (int),
 894                 .mode           = 0644,
 895                 .proc_handler   = proc_dointvec,
 896         },
 897 #endif
 898 #ifdef CONFIG_RT_MUTEXES
 899         {
 900                 .procname       = "max_lock_depth",
 901                 .data           = &max_lock_depth,
 902                 .maxlen         = sizeof(int),
 903                 .mode           = 0644,
 904                 .proc_handler   = proc_dointvec,
 905         },
 906 #endif
 907         {
 908                 .procname       = "poweroff_cmd",
 909                 .data           = &poweroff_cmd,
 910                 .maxlen         = POWEROFF_CMD_PATH_LEN,
 911                 .mode           = 0644,
 912                 .proc_handler   = proc_dostring,
 913         },
 914 #ifdef CONFIG_KEYS
 915         {
 916                 .procname       = "keys",
 917                 .mode           = 0555,
 918                 .child          = key_sysctls,
 919         },
 920 #endif
 921 #ifdef CONFIG_RCU_TORTURE_TEST
 922         {
 923                 .procname       = "rcutorture_runnable",
 924                 .data           = &rcutorture_runnable,
 925                 .maxlen         = sizeof(int),
 926                 .mode           = 0644,
 927                 .proc_handler   = proc_dointvec,
 928         },
 929 #endif
 930 #ifdef CONFIG_PERF_EVENTS
 931         {
 932                 .procname       = "perf_event_paranoid",
 933                 .data           = &sysctl_perf_event_paranoid,
 934                 .maxlen         = sizeof(sysctl_perf_event_paranoid),
 935                 .mode           = 0644,
 936                 .proc_handler   = proc_dointvec,
 937         },
 938         {
 939                 .procname       = "perf_event_mlock_kb",
 940                 .data           = &sysctl_perf_event_mlock,
 941                 .maxlen         = sizeof(sysctl_perf_event_mlock),
 942                 .mode           = 0644,
 943                 .proc_handler   = proc_dointvec,
 944         },
 945         {
 946                 .procname       = "perf_event_max_sample_rate",
 947                 .data           = &sysctl_perf_event_sample_rate,
 948                 .maxlen         = sizeof(sysctl_perf_event_sample_rate),
 949                 .mode           = 0644,
 950                 .proc_handler   = perf_proc_update_handler,
 951         },
 952 #endif
 953 #ifdef CONFIG_KMEMCHECK
 954         {
 955                 .procname       = "kmemcheck",
 956                 .data           = &kmemcheck_enabled,
 957                 .maxlen         = sizeof(int),
 958                 .mode           = 0644,
 959                 .proc_handler   = proc_dointvec,
 960         },
 961 #endif
 962 #ifdef CONFIG_BLOCK
 963         {
 964                 .procname       = "blk_iopoll",
 965                 .data           = &blk_iopoll_enabled,
 966                 .maxlen         = sizeof(int),
 967                 .mode           = 0644,
 968                 .proc_handler   = proc_dointvec,
 969         },
 970 #endif
 971         { }
 972 };
 973
 974 static struct ctl_table vm_table[] = {
 975         {
 976                 .procname       = "overcommit_memory",
 977                 .data           = &sysctl_overcommit_memory,
 978                 .maxlen         = sizeof(sysctl_overcommit_memory),
 979                 .mode           = 0644,
 980                 .proc_handler   = proc_dointvec_minmax,
 981                 .extra1         = &zero,
 982                 .extra2         = &two,
 983         },
 984         {
 985                 .procname       = "panic_on_oom",
 986                 .data           = &sysctl_panic_on_oom,
 987                 .maxlen         = sizeof(sysctl_panic_on_oom),
 988                 .mode           = 0644,
 989                 .proc_handler   = proc_dointvec_minmax,
 990                 .extra1         = &zero,
 991                 .extra2         = &two,
 992         },
 993         {
 994                 .procname       = "oom_kill_allocating_task",
 995                 .data           = &sysctl_oom_kill_allocating_task,
 996                 .maxlen         = sizeof(sysctl_oom_kill_allocating_task),
 997                 .mode           = 0644,
 998                 .proc_handler   = proc_dointvec,
 999         },
1000         {
1001                 .procname       = "oom_dump_tasks",
1002                 .data           = &sysctl_oom_dump_tasks,
1003                 .maxlen         = sizeof(sysctl_oom_dump_tasks),
1004                 .mode           = 0644,
1005                 .proc_handler   = proc_dointvec,
1006         },
1007         {
1008                 .procname       = "overcommit_ratio",
1009                 .data           = &sysctl_overcommit_ratio,
1010                 .maxlen         = sizeof(sysctl_overcommit_ratio),
1011                 .mode           = 0644,
1012                 .proc_handler   = proc_dointvec,
1013         },
1014         {
1015                 .procname       = "page-cluster",
1016                 .data           = &page_cluster,
1017                 .maxlen         = sizeof(int),
1018                 .mode           = 0644,
1019                 .proc_handler   = proc_dointvec_minmax,
1020                 .extra1         = &zero,
1021         },
1022         {
1023                 .procname       = "dirty_background_ratio",
1024                 .data           = &dirty_background_ratio,
1025                 .maxlen         = sizeof(dirty_background_ratio),
1026                 .mode           = 0644,
1027                 .proc_handler   = dirty_background_ratio_handler,
1028                 .extra1         = &zero,
1029                 .extra2         = &one_hundred,
1030         },
1031         {
1032                 .procname       = "dirty_background_bytes",
1033                 .data           = &dirty_background_bytes,
1034                 .maxlen         = sizeof(dirty_background_bytes),
1035                 .mode           = 0644,
1036                 .proc_handler   = dirty_background_bytes_handler,
1037                 .extra1         = &one_ul,
1038         },
1039         {
1040                 .procname       = "dirty_ratio",
1041                 .data           = &vm_dirty_ratio,
1042                 .maxlen         = sizeof(vm_dirty_ratio),
1043                 .mode           = 0644,
1044                 .proc_handler   = dirty_ratio_handler,
1045                 .extra1         = &zero,
1046                 .extra2         = &one_hundred,
1047         },
1048         {
1049                 .procname       = "dirty_bytes",
1050                 .data           = &vm_dirty_bytes,
1051                 .maxlen         = sizeof(vm_dirty_bytes),
1052                 .mode           = 0644,
1053                 .proc_handler   = dirty_bytes_handler,
1054                 .extra1         = &dirty_bytes_min,
1055         },
1056         {
1057                 .procname       = "dirty_writeback_centisecs",
1058                 .data           = &dirty_writeback_interval,
1059                 .maxlen         = sizeof(dirty_writeback_interval),
1060                 .mode           = 0644,
1061                 .proc_handler   = dirty_writeback_centisecs_handler,
1062         },
1063         {
1064                 .procname       = "dirty_expire_centisecs",
1065                 .data           = &dirty_expire_interval,
1066                 .maxlen         = sizeof(dirty_expire_interval),
1067                 .mode           = 0644,
1068                 .proc_handler   = proc_dointvec_minmax,
1069                 .extra1         = &zero,
1070         },
1071         {
1072                 .procname       = "nr_pdflush_threads",
1073                 .data           = &nr_pdflush_threads,
1074                 .maxlen         = sizeof nr_pdflush_threads,
1075                 .mode           = 0444 /* read-only*/,
1076                 .proc_handler   = proc_dointvec,
1077         },
1078         {
1079                 .procname       = "swappiness",
1080                 .data           = &vm_swappiness,
1081                 .maxlen         = sizeof(vm_swappiness),
1082                 .mode           = 0644,
1083                 .proc_handler   = proc_dointvec_minmax,
1084                 .extra1         = &zero,
1085                 .extra2         = &one_hundred,
1086         },
1087 #ifdef CONFIG_HUGETLB_PAGE
1088         {
1089                 .procname       = "nr_hugepages",
1090                 .data           = NULL,
1091                 .maxlen         = sizeof(unsigned long),
1092                 .mode           = 0644,
1093                 .proc_handler   = hugetlb_sysctl_handler,
1094                 .extra1         = (void *)&hugetlb_zero,
1095                 .extra2         = (void *)&hugetlb_infinity,
1096         },
1097 #ifdef CONFIG_NUMA
1098         {
1099                 .procname       = "nr_hugepages_mempolicy",
1100                 .data           = NULL,
1101                 .maxlen         = sizeof(unsigned long),
1102                 .mode           = 0644,
1103                 .proc_handler   = &hugetlb_mempolicy_sysctl_handler,
1104                 .extra1         = (void *)&hugetlb_zero,
1105                 .extra2         = (void *)&hugetlb_infinity,
1106         },
1107 #endif
1108          {
1109                 .procname       = "hugetlb_shm_group",
1110                 .data           = &sysctl_hugetlb_shm_group,
1111                 .maxlen         = sizeof(gid_t),
1112                 .mode           = 0644,
1113                 .proc_handler   = proc_dointvec,
1114          },
1115          {
1116                 .procname       = "hugepages_treat_as_movable",
1117                 .data           = &hugepages_treat_as_movable,
1118                 .maxlen         = sizeof(int),
1119                 .mode           = 0644,
1120                 .proc_handler   = hugetlb_treat_movable_handler,
1121         },
1122         {
1123                 .procname       = "nr_overcommit_hugepages",
1124                 .data           = NULL,
1125                 .maxlen         = sizeof(unsigned long),
1126                 .mode           = 0644,
1127                 .proc_handler   = hugetlb_overcommit_handler,
1128                 .extra1         = (void *)&hugetlb_zero,
1129                 .extra2         = (void *)&hugetlb_infinity,
1130         },
1131 #endif
1132         {
1133                 .procname       = "lowmem_reserve_ratio",
1134                 .data           = &sysctl_lowmem_reserve_ratio,
1135                 .maxlen         = sizeof(sysctl_lowmem_reserve_ratio),
1136                 .mode           = 0644,
1137                 .proc_handler   = lowmem_reserve_ratio_sysctl_handler,
1138         },
1139         {
1140                 .procname       = "drop_caches",
1141                 .data           = &sysctl_drop_caches,
1142                 .maxlen         = sizeof(int),
1143                 .mode           = 0644,
1144                 .proc_handler   = drop_caches_sysctl_handler,
1145                 .extra1         = &one,
1146                 .extra2         = &three,
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
1342 #if defined(CONFIG_BINFMT_MISC) || defined(CONFIG_BINFMT_MISC_MODULE)
1343 static struct ctl_table binfmt_misc_table[] = {
1344         { }
1345 };
1346 #endif
1347
1348 static struct ctl_table fs_table[] = {
1349         {
1350                 .procname       = "inode-nr",
1351                 .data           = &inodes_stat,
1352                 .maxlen         = 2*sizeof(int),
1353                 .mode           = 0444,
1354                 .proc_handler   = proc_nr_inodes,
1355         },
1356         {
1357                 .procname       = "inode-state",
1358                 .data           = &inodes_stat,
1359                 .maxlen         = 7*sizeof(int),
1360                 .mode           = 0444,
1361                 .proc_handler   = proc_nr_inodes,
1362         },
1363         {
1364                 .procname       = "file-nr",
1365                 .data           = &files_stat,
1366                 .maxlen         = sizeof(files_stat),
1367                 .mode           = 0444,
1368                 .proc_handler   = proc_nr_files,
1369         },
1370         {
1371                 .procname       = "file-max",
1372                 .data           = &files_stat.max_files,
1373                 .maxlen         = sizeof(files_stat.max_files),
1374                 .mode           = 0644,
1375                 .proc_handler   = proc_doulongvec_minmax,
1376         },
1377         {
1378                 .procname       = "nr_open",
1379                 .data           = &sysctl_nr_open,
1380                 .maxlen         = sizeof(int),
1381                 .mode           = 0644,
1382                 .proc_handler   = proc_dointvec_minmax,
1383                 .extra1         = &sysctl_nr_open_min,
1384                 .extra2         = &sysctl_nr_open_max,
1385         },
1386         {
1387                 .procname       = "dentry-state",
1388                 .data           = &dentry_stat,
1389                 .maxlen         = 6*sizeof(int),
1390                 .mode           = 0444,
1391                 .proc_handler   = proc_nr_dentry,
1392         },
1393         {
1394                 .procname       = "overflowuid",
1395                 .data           = &fs_overflowuid,
1396                 .maxlen         = sizeof(int),
1397                 .mode           = 0644,
1398                 .proc_handler   = proc_dointvec_minmax,
1399                 .extra1         = &minolduid,
1400                 .extra2         = &maxolduid,
1401         },
1402         {
1403                 .procname       = "overflowgid",
1404                 .data           = &fs_overflowgid,
1405                 .maxlen         = sizeof(int),
1406                 .mode           = 0644,
1407                 .proc_handler   = proc_dointvec_minmax,
1408                 .extra1         = &minolduid,
1409                 .extra2         = &maxolduid,
1410         },
1411 #ifdef CONFIG_FILE_LOCKING
1412         {
1413                 .procname       = "leases-enable",
1414                 .data           = &leases_enable,
1415                 .maxlen         = sizeof(int),
1416                 .mode           = 0644,
1417                 .proc_handler   = proc_dointvec,
1418         },
1419 #endif
1420 #ifdef CONFIG_DNOTIFY
1421         {
1422                 .procname       = "dir-notify-enable",
1423                 .data           = &dir_notify_enable,
1424                 .maxlen         = sizeof(int),
1425                 .mode           = 0644,
1426                 .proc_handler   = proc_dointvec,
1427         },
1428 #endif
1429 #ifdef CONFIG_MMU
1430 #ifdef CONFIG_FILE_LOCKING
1431         {
1432                 .procname       = "lease-break-time",
1433                 .data           = &lease_break_time,
1434                 .maxlen         = sizeof(int),
1435                 .mode           = 0644,
1436                 .proc_handler   = proc_dointvec,
1437         },
1438 #endif
1439 #ifdef CONFIG_AIO
1440         {
1441                 .procname       = "aio-nr",
1442                 .data           = &aio_nr,
1443                 .maxlen         = sizeof(aio_nr),
1444                 .mode           = 0444,
1445                 .proc_handler   = proc_doulongvec_minmax,
1446         },
1447         {
1448                 .procname       = "aio-max-nr",
1449                 .data           = &aio_max_nr,
1450                 .maxlen         = sizeof(aio_max_nr),
1451                 .mode           = 0644,
1452                 .proc_handler   = proc_doulongvec_minmax,
1453         },
1454 #endif /* CONFIG_AIO */
1455 #ifdef CONFIG_INOTIFY_USER
1456         {
1457                 .procname       = "inotify",
1458                 .mode           = 0555,
1459                 .child          = inotify_table,
1460         },
1461 #endif
1462 #ifdef CONFIG_EPOLL
1463         {
1464                 .procname       = "epoll",
1465                 .mode           = 0555,
1466                 .child          = epoll_table,
1467         },
1468 #endif
1469 #endif
1470         {
1471                 .procname       = "suid_dumpable",
1472                 .data           = &suid_dumpable,
1473                 .maxlen         = sizeof(int),
1474                 .mode           = 0644,
1475                 .proc_handler   = proc_dointvec_minmax,
1476                 .extra1         = &zero,
1477                 .extra2         = &two,
1478         },
1479 #if defined(CONFIG_BINFMT_MISC) || defined(CONFIG_BINFMT_MISC_MODULE)
1480         {
1481                 .procname       = "binfmt_misc",
1482                 .mode           = 0555,
1483                 .child          = binfmt_misc_table,
1484         },
1485 #endif
1486         {
1487                 .procname       = "pipe-max-size",
1488                 .data           = &pipe_max_size,
1489                 .maxlen         = sizeof(int),
1490                 .mode           = 0644,
1491                 .proc_handler   = &pipe_proc_fn,
1492                 .extra1         = &pipe_min_size,
1493         },
1494         { }
1495 };
1496
1497 static struct ctl_table debug_table[] = {
1498 #if defined(CONFIG_X86) || defined(CONFIG_PPC) || defined(CONFIG_SPARC) || \
1499     defined(CONFIG_S390)
1500         {
1501                 .procname       = "exception-trace",
1502                 .data           = &show_unhandled_signals,
1503                 .maxlen         = sizeof(int),
1504                 .mode           = 0644,
1505                 .proc_handler   = proc_dointvec
1506         },
1507 #endif
1508 #if defined(CONFIG_OPTPROBES)
1509         {
1510                 .procname       = "kprobes-optimization",
1511                 .data           = &sysctl_kprobes_optimization,
1512                 .maxlen         = sizeof(int),
1513                 .mode           = 0644,
1514                 .proc_handler   = proc_kprobes_optimization_handler,
1515                 .extra1         = &zero,
1516                 .extra2         = &one,
1517         },
1518 #endif
1519         { }
1520 };
1521
1522 static struct ctl_table dev_table[] = {
1523         { }
1524 };
1525
1526 static DEFINE_SPINLOCK(sysctl_lock);
1527
1528 /* called under sysctl_lock */
1529 static int use_table(struct ctl_table_header *p)
1530 {
1531         if (unlikely(p->unregistering))
1532                 return 0;
1533         p->used++;
1534         return 1;
1535 }
1536
1537 /* called under sysctl_lock */
1538 static void unuse_table(struct ctl_table_header *p)
1539 {
1540         if (!--p->used)
1541                 if (unlikely(p->unregistering))
1542                         complete(p->unregistering);
1543 }
1544
1545 /* called under sysctl_lock, will reacquire if has to wait */
1546 static void start_unregistering(struct ctl_table_header *p)
1547 {
1548         /*
1549          * if p->used is 0, nobody will ever touch that entry again;
1550          * we'll eliminate all paths to it before dropping sysctl_lock
1551          */
1552         if (unlikely(p->used)) {
1553                 struct completion wait;
1554                 init_completion(&wait);
1555                 p->unregistering = &wait;
1556                 spin_unlock(&sysctl_lock);
1557                 wait_for_completion(&wait);
1558                 spin_lock(&sysctl_lock);
1559         } else {
1560                 /* anything non-NULL; we'll never dereference it */
1561                 p->unregistering = ERR_PTR(-EINVAL);
1562         }
1563         /*
1564          * do not remove from the list until nobody holds it; walking the
1565          * list in do_sysctl() relies on that.
1566          */
1567         list_del_init(&p->ctl_entry);
1568 }
1569
1570 void sysctl_head_get(struct ctl_table_header *head)
1571 {
1572         spin_lock(&sysctl_lock);
1573         head->count++;
1574         spin_unlock(&sysctl_lock);
1575 }
1576
1577 static void free_head(struct rcu_head *rcu)
1578 {
1579         kfree(container_of(rcu, struct ctl_table_header, rcu));
1580 }
1581
1582 void sysctl_head_put(struct ctl_table_header *head)
1583 {
1584         spin_lock(&sysctl_lock);
1585         if (!--head->count)
1586                 call_rcu(&head->rcu, free_head);
1587         spin_unlock(&sysctl_lock);
1588 }
1589
1590 struct ctl_table_header *sysctl_head_grab(struct ctl_table_header *head)
1591 {
1592         if (!head)
1593                 BUG();
1594         spin_lock(&sysctl_lock);
1595         if (!use_table(head))
1596                 head = ERR_PTR(-ENOENT);
1597         spin_unlock(&sysctl_lock);
1598         return head;
1599 }
1600
1601 void sysctl_head_finish(struct ctl_table_header *head)
1602 {
1603         if (!head)
1604                 return;
1605         spin_lock(&sysctl_lock);
1606         unuse_table(head);
1607         spin_unlock(&sysctl_lock);
1608 }
1609
1610 static struct ctl_table_set *
1611 lookup_header_set(struct ctl_table_root *root, struct nsproxy *namespaces)
1612 {
1613         struct ctl_table_set *set = &root->default_set;
1614         if (root->lookup)
1615                 set = root->lookup(root, namespaces);
1616         return set;
1617 }
1618
1619 static struct list_head *
1620 lookup_header_list(struct ctl_table_root *root, struct nsproxy *namespaces)
1621 {
1622         struct ctl_table_set *set = lookup_header_set(root, namespaces);
1623         return &set->list;
1624 }
1625
1626 struct ctl_table_header *__sysctl_head_next(struct nsproxy *namespaces,
1627                                             struct ctl_table_header *prev)
1628 {
1629         struct ctl_table_root *root;
1630         struct list_head *header_list;
1631         struct ctl_table_header *head;
1632         struct list_head *tmp;
1633
1634         spin_lock(&sysctl_lock);
1635         if (prev) {
1636                 head = prev;
1637                 tmp = &prev->ctl_entry;
1638                 unuse_table(prev);
1639                 goto next;
1640         }
1641         tmp = &root_table_header.ctl_entry;
1642         for (;;) {
1643                 head = list_entry(tmp, struct ctl_table_header, ctl_entry);
1644
1645                 if (!use_table(head))
1646                         goto next;
1647                 spin_unlock(&sysctl_lock);
1648                 return head;
1649         next:
1650                 root = head->root;
1651                 tmp = tmp->next;
1652                 header_list = lookup_header_list(root, namespaces);
1653                 if (tmp != header_list)
1654                         continue;
1655
1656                 do {
1657                         root = list_entry(root->root_list.next,
1658                                         struct ctl_table_root, root_list);
1659                         if (root == &sysctl_table_root)
1660                                 goto out;
1661                         header_list = lookup_header_list(root, namespaces);
1662                 } while (list_empty(header_list));
1663                 tmp = header_list->next;
1664         }
1665 out:
1666         spin_unlock(&sysctl_lock);
1667         return NULL;
1668 }
1669
1670 struct ctl_table_header *sysctl_head_next(struct ctl_table_header *prev)
1671 {
1672         return __sysctl_head_next(current->nsproxy, prev);
1673 }
1674
1675 void register_sysctl_root(struct ctl_table_root *root)
1676 {
1677         spin_lock(&sysctl_lock);
1678         list_add_tail(&root->root_list, &sysctl_table_root.root_list);
1679         spin_unlock(&sysctl_lock);
1680 }
1681
1682 /*
1683  * sysctl_perm does NOT grant the superuser all rights automatically, because
1684  * some sysctl variables are readonly even to root.
1685  */
1686
1687 static int test_perm(int mode, int op)
1688 {
1689         if (!current_euid())
1690                 mode >>= 6;
1691         else if (in_egroup_p(0))
1692                 mode >>= 3;
1693         if ((op & ~mode & (MAY_READ|MAY_WRITE|MAY_EXEC)) == 0)
1694                 return 0;
1695         return -EACCES;
1696 }
1697
1698 int sysctl_perm(struct ctl_table_root *root, struct ctl_table *table, int op)
1699 {
1700         int mode;
1701
1702         if (root->permissions)
1703                 mode = root->permissions(root, current->nsproxy, table);
1704         else
1705                 mode = table->mode;
1706
1707         return test_perm(mode, op);
1708 }
1709
1710 static void sysctl_set_parent(struct ctl_table *parent, struct ctl_table *table)
1711 {
1712         for (; table->procname; table++) {
1713                 table->parent = parent;
1714                 if (table->child)
1715                         sysctl_set_parent(table, table->child);
1716         }
1717 }
1718
1719 static __init int sysctl_init(void)
1720 {
1721         sysctl_set_parent(NULL, root_table);
1722 #ifdef CONFIG_SYSCTL_SYSCALL_CHECK
1723         sysctl_check_table(current->nsproxy, root_table);
1724 #endif
1725         return 0;
1726 }
1727
1728 core_initcall(sysctl_init);
1729
1730 static struct ctl_table *is_branch_in(struct ctl_table *branch,
1731                                       struct ctl_table *table)
1732 {
1733         struct ctl_table *p;
1734         const char *s = branch->procname;
1735
1736         /* branch should have named subdirectory as its first element */
1737         if (!s || !branch->child)
1738                 return NULL;
1739
1740         /* ... and nothing else */
1741         if (branch[1].procname)
1742                 return NULL;
1743
1744         /* table should contain subdirectory with the same name */
1745         for (p = table; p->procname; p++) {
1746                 if (!p->child)
1747                         continue;
1748                 if (p->procname && strcmp(p->procname, s) == 0)
1749                         return p;
1750         }
1751         return NULL;
1752 }
1753
1754 /* see if attaching q to p would be an improvement */
1755 static void try_attach(struct ctl_table_header *p, struct ctl_table_header *q)
1756 {
1757         struct ctl_table *to = p->ctl_table, *by = q->ctl_table;
1758         struct ctl_table *next;
1759         int is_better = 0;
1760         int not_in_parent = !p->attached_by;
1761
1762         while ((next = is_branch_in(by, to)) != NULL) {
1763                 if (by == q->attached_by)
1764                         is_better = 1;
1765                 if (to == p->attached_by)
1766                         not_in_parent = 1;
1767                 by = by->child;
1768                 to = next->child;
1769         }
1770
1771         if (is_better && not_in_parent) {
1772                 q->attached_by = by;
1773                 q->attached_to = to;
1774                 q->parent = p;
1775         }
1776 }
1777
1778 /**
1779  * __register_sysctl_paths - register a sysctl hierarchy
1780  * @root: List of sysctl headers to register on
1781  * @namespaces: Data to compute which lists of sysctl entries are visible
1782  * @path: The path to the directory the sysctl table is in.
1783  * @table: the top-level table structure
1784  *
1785  * Register a sysctl table hierarchy. @table should be a filled in ctl_table
1786  * array. A completely 0 filled entry terminates the table.
1787  *
1788  * The members of the &struct ctl_table structure are used as follows:
1789  *
1790  * procname - the name of the sysctl file under /proc/sys. Set to %NULL to not
1791  *            enter a sysctl file
1792  *
1793  * data - a pointer to data for use by proc_handler
1794  *
1795  * maxlen - the maximum size in bytes of the data
1796  *
1797  * mode - the file permissions for the /proc/sys file, and for sysctl(2)
1798  *
1799  * child - a pointer to the child sysctl table if this entry is a directory, or
1800  *         %NULL.
1801  *
1802  * proc_handler - the text handler routine (described below)
1803  *
1804  * de - for internal use by the sysctl routines
1805  *
1806  * extra1, extra2 - extra pointers usable by the proc handler routines
1807  *
1808  * Leaf nodes in the sysctl tree will be represented by a single file
1809  * under /proc; non-leaf nodes will be represented by directories.
1810  *
1811  * sysctl(2) can automatically manage read and write requests through
1812  * the sysctl table.  The data and maxlen fields of the ctl_table
1813  * struct enable minimal validation of the values being written to be
1814  * performed, and the mode field allows minimal authentication.
1815  *
1816  * There must be a proc_handler routine for any terminal nodes
1817  * mirrored under /proc/sys (non-terminals are handled by a built-in
1818  * directory handler).  Several default handlers are available to
1819  * cover common cases -
1820  *
1821  * proc_dostring(), proc_dointvec(), proc_dointvec_jiffies(),
1822  * proc_dointvec_userhz_jiffies(), proc_dointvec_minmax(),
1823  * proc_doulongvec_ms_jiffies_minmax(), proc_doulongvec_minmax()
1824  *
1825  * It is the handler's job to read the input buffer from user memory
1826  * and process it. The handler should return 0 on success.
1827  *
1828  * This routine returns %NULL on a failure to register, and a pointer
1829  * to the table header on success.
1830  */
1831 struct ctl_table_header *__register_sysctl_paths(
1832         struct ctl_table_root *root,
1833         struct nsproxy *namespaces,
1834         const struct ctl_path *path, struct ctl_table *table)
1835 {
1836         struct ctl_table_header *header;
1837         struct ctl_table *new, **prevp;
1838         unsigned int n, npath;
1839         struct ctl_table_set *set;
1840
1841         /* Count the path components */
1842         for (npath = 0; path[npath].procname; ++npath)
1843                 ;
1844
1845         /*
1846          * For each path component, allocate a 2-element ctl_table array.
1847          * The first array element will be filled with the sysctl entry
1848          * for this, the second will be the sentinel (procname == 0).
1849          *
1850          * We allocate everything in one go so that we don't have to
1851          * worry about freeing additional memory in unregister_sysctl_table.
1852          */
1853         header = kzalloc(sizeof(struct ctl_table_header) +
1854                          (2 * npath * sizeof(struct ctl_table)), GFP_KERNEL);
1855         if (!header)
1856                 return NULL;
1857
1858         new = (struct ctl_table *) (header + 1);
1859
1860         /* Now connect the dots */
1861         prevp = &header->ctl_table;
1862         for (n = 0; n < npath; ++n, ++path) {
1863                 /* Copy the procname */
1864                 new->procname = path->procname;
1865                 new->mode     = 0555;
1866
1867                 *prevp = new;
1868                 prevp = &new->child;
1869
1870                 new += 2;
1871         }
1872         *prevp = table;
1873         header->ctl_table_arg = table;
1874
1875         INIT_LIST_HEAD(&header->ctl_entry);
1876         header->used = 0;
1877         header->unregistering = NULL;
1878         header->root = root;
1879         sysctl_set_parent(NULL, header->ctl_table);
1880         header->count = 1;
1881 #ifdef CONFIG_SYSCTL_SYSCALL_CHECK
1882         if (sysctl_check_table(namespaces, header->ctl_table)) {
1883                 kfree(header);
1884                 return NULL;
1885         }
1886 #endif
1887         spin_lock(&sysctl_lock);
1888         header->set = lookup_header_set(root, namespaces);
1889         header->attached_by = header->ctl_table;
1890         header->attached_to = root_table;
1891         header->parent = &root_table_header;
1892         for (set = header->set; set; set = set->parent) {
1893                 struct ctl_table_header *p;
1894                 list_for_each_entry(p, &set->list, ctl_entry) {
1895                         if (p->unregistering)
1896                                 continue;
1897                         try_attach(p, header);
1898                 }
1899         }
1900         header->parent->count++;
1901         list_add_tail(&header->ctl_entry, &header->set->list);
1902         spin_unlock(&sysctl_lock);
1903
1904         return header;
1905 }
1906
1907 /**
1908  * register_sysctl_table_path - register a sysctl table hierarchy
1909  * @path: The path to the directory the sysctl table is in.
1910  * @table: the top-level table structure
1911  *
1912  * Register a sysctl table hierarchy. @table should be a filled in ctl_table
1913  * array. A completely 0 filled entry terminates the table.
1914  *
1915  * See __register_sysctl_paths for more details.
1916  */
1917 struct ctl_table_header *register_sysctl_paths(const struct ctl_path *path,
1918                                                 struct ctl_table *table)
1919 {
1920         return __register_sysctl_paths(&sysctl_table_root, current->nsproxy,
1921                                         path, table);
1922 }
1923
1924 /**
1925  * register_sysctl_table - register a sysctl table hierarchy
1926  * @table: the top-level table structure
1927  *
1928  * Register a sysctl table hierarchy. @table should be a filled in ctl_table
1929  * array. A completely 0 filled entry terminates the table.
1930  *
1931  * See register_sysctl_paths for more details.
1932  */
1933 struct ctl_table_header *register_sysctl_table(struct ctl_table *table)
1934 {
1935         static const struct ctl_path null_path[] = { {} };
1936
1937         return register_sysctl_paths(null_path, table);
1938 }
1939
1940 /**
1941  * unregister_sysctl_table - unregister a sysctl table hierarchy
1942  * @header: the header returned from register_sysctl_table
1943  *
1944  * Unregisters the sysctl table and all children. proc entries may not
1945  * actually be removed until they are no longer used by anyone.
1946  */
1947 void unregister_sysctl_table(struct ctl_table_header * header)
1948 {
1949         might_sleep();
1950
1951         if (header == NULL)
1952                 return;
1953
1954         spin_lock(&sysctl_lock);
1955         start_unregistering(header);
1956         if (!--header->parent->count) {
1957                 WARN_ON(1);
1958                 call_rcu(&header->parent->rcu, free_head);
1959         }
1960         if (!--header->count)
1961                 call_rcu(&header->rcu, free_head);
1962         spin_unlock(&sysctl_lock);
1963 }
1964
1965 int sysctl_is_seen(struct ctl_table_header *p)
1966 {
1967         struct ctl_table_set *set = p->set;
1968         int res;
1969         spin_lock(&sysctl_lock);
1970         if (p->unregistering)
1971                 res = 0;
1972         else if (!set->is_seen)
1973                 res = 1;
1974         else
1975                 res = set->is_seen(set);
1976         spin_unlock(&sysctl_lock);
1977         return res;
1978 }
1979
1980 void setup_sysctl_set(struct ctl_table_set *p,
1981         struct ctl_table_set *parent,
1982         int (*is_seen)(struct ctl_table_set *))
1983 {
1984         INIT_LIST_HEAD(&p->list);
1985         p->parent = parent ? parent : &sysctl_table_root.default_set;
1986         p->is_seen = is_seen;
1987 }
1988
1989 #else /* !CONFIG_SYSCTL */
1990 struct ctl_table_header *register_sysctl_table(struct ctl_table * table)
1991 {
1992         return NULL;
1993 }
1994
1995 struct ctl_table_header *register_sysctl_paths(const struct ctl_path *path,
1996                                                     struct ctl_table *table)
1997 {
1998         return NULL;
1999 }
2000
2001 void unregister_sysctl_table(struct ctl_table_header * table)
2002 {
2003 }
2004
2005 void setup_sysctl_set(struct ctl_table_set *p,
2006         struct ctl_table_set *parent,
2007         int (*is_seen)(struct ctl_table_set *))
2008 {
2009 }
2010
2011 void sysctl_head_put(struct ctl_table_header *head)
2012 {
2013 }
2014
2015 #endif /* CONFIG_SYSCTL */
2016
2017 /*
2018  * /proc/sys support
2019  */
2020
2021 #ifdef CONFIG_PROC_SYSCTL
2022
2023 static int _proc_do_string(void* data, int maxlen, int write,
2024                            void __user *buffer,
2025                            size_t *lenp, loff_t *ppos)
2026 {
2027         size_t len;
2028         char __user *p;
2029         char c;
2030
2031         if (!data || !maxlen || !*lenp) {
2032                 *lenp = 0;
2033                 return 0;
2034         }
2035
2036         if (write) {
2037                 len = 0;
2038                 p = buffer;
2039                 while (len < *lenp) {
2040                         if (get_user(c, p++))
2041                                 return -EFAULT;
2042                         if (c == 0 || c == '\n')
2043                                 break;
2044                         len++;
2045                 }
2046                 if (len >= maxlen)
2047                         len = maxlen-1;
2048                 if(copy_from_user(data, buffer, len))
2049                         return -EFAULT;
2050                 ((char *) data)[len] = 0;
2051                 *ppos += *lenp;
2052         } else {
2053                 len = strlen(data);
2054                 if (len > maxlen)
2055                         len = maxlen;
2056
2057                 if (*ppos > len) {
2058                         *lenp = 0;
2059                         return 0;
2060                 }
2061
2062                 data += *ppos;
2063                 len  -= *ppos;
2064
2065                 if (len > *lenp)
2066                         len = *lenp;
2067                 if (len)
2068                         if(copy_to_user(buffer, data, len))
2069                                 return -EFAULT;
2070                 if (len < *lenp) {
2071                         if(put_user('\n', ((char __user *) buffer) + len))
2072                                 return -EFAULT;
2073                         len++;
2074                 }
2075                 *lenp = len;
2076                 *ppos += len;
2077         }
2078         return 0;
2079 }
2080
2081 /**
2082  * proc_dostring - read a string sysctl
2083  * @table: the sysctl table
2084  * @write: %TRUE if this is a write to the sysctl file
2085  * @buffer: the user buffer
2086  * @lenp: the size of the user buffer
2087  * @ppos: file position
2088  *
2089  * Reads/writes a string from/to the user buffer. If the kernel
2090  * buffer provided is not large enough to hold the string, the
2091  * string is truncated. The copied string is %NULL-terminated.
2092  * If the string is being read by the user process, it is copied
2093  * and a newline '\n' is added. It is truncated if the buffer is
2094  * not large enough.
2095  *
2096  * Returns 0 on success.
2097  */
2098 int proc_dostring(struct ctl_table *table, int write,
2099                   void __user *buffer, size_t *lenp, loff_t *ppos)
2100 {
2101         return _proc_do_string(table->data, table->maxlen, write,
2102                                buffer, lenp, ppos);
2103 }
2104
2105 static size_t proc_skip_spaces(char **buf)
2106 {
2107         size_t ret;
2108         char *tmp = skip_spaces(*buf);
2109         ret = tmp - *buf;
2110         *buf = tmp;
2111         return ret;
2112 }
2113
2114 static void proc_skip_char(char **buf, size_t *size, const char v)
2115 {
2116         while (*size) {
2117                 if (**buf != v)
2118                         break;
2119                 (*size)--;
2120                 (*buf)++;
2121         }
2122 }
2123
2124 #define TMPBUFLEN 22
2125 /**
2126  * proc_get_long - reads an ASCII formatted integer from a user buffer
2127  *
2128  * @buf: a kernel buffer
2129  * @size: size of the kernel buffer
2130  * @val: this is where the number will be stored
2131  * @neg: set to %TRUE if number is negative
2132  * @perm_tr: a vector which contains the allowed trailers
2133  * @perm_tr_len: size of the perm_tr vector
2134  * @tr: pointer to store the trailer character
2135  *
2136  * In case of success %0 is returned and @buf and @size are updated with
2137  * the amount of bytes read. If @tr is non-NULL and a trailing
2138  * character exists (size is non-zero after returning from this
2139  * function), @tr is updated with the trailing character.
2140  */
2141 static int proc_get_long(char **buf, size_t *size,
2142                           unsigned long *val, bool *neg,
2143                           const char *perm_tr, unsigned perm_tr_len, char *tr)
2144 {
2145         int len;
2146         char *p, tmp[TMPBUFLEN];
2147
2148         if (!*size)
2149                 return -EINVAL;
2150
2151         len = *size;
2152         if (len > TMPBUFLEN - 1)
2153                 len = TMPBUFLEN - 1;
2154
2155         memcpy(tmp, *buf, len);
2156
2157         tmp[len] = 0;
2158         p = tmp;
2159         if (*p == '-' && *size > 1) {
2160                 *neg = true;
2161                 p++;
2162         } else
2163                 *neg = false;
2164         if (!isdigit(*p))
2165                 return -EINVAL;
2166
2167         *val = simple_strtoul(p, &p, 0);
2168
2169         len = p - tmp;
2170
2171         /* We don't know if the next char is whitespace thus we may accept
2172          * invalid integers (e.g. 1234...a) or two integers instead of one
2173          * (e.g. 123...1). So lets not allow such large numbers. */
2174         if (len == TMPBUFLEN - 1)
2175                 return -EINVAL;
2176
2177         if (len < *size && perm_tr_len && !memchr(perm_tr, *p, perm_tr_len))
2178                 return -EINVAL;
2179
2180         if (tr && (len < *size))
2181                 *tr = *p;
2182
2183         *buf += len;
2184         *size -= len;
2185
2186         return 0;
2187 }
2188
2189 /**
2190  * proc_put_long - converts an integer to a decimal ASCII formatted string
2191  *
2192  * @buf: the user buffer
2193  * @size: the size of the user buffer
2194  * @val: the integer to be converted
2195  * @neg: sign of the number, %TRUE for negative
2196  *
2197  * In case of success %0 is returned and @buf and @size are updated with
2198  * the amount of bytes written.
2199  */
2200 static int proc_put_long(void __user **buf, size_t *size, unsigned long val,
2201                           bool neg)
2202 {
2203         int len;
2204         char tmp[TMPBUFLEN], *p = tmp;
2205
2206         sprintf(p, "%s%lu", neg ? "-" : "", val);
2207         len = strlen(tmp);
2208         if (len > *size)
2209                 len = *size;
2210         if (copy_to_user(*buf, tmp, len))
2211                 return -EFAULT;
2212         *size -= len;
2213         *buf += len;
2214         return 0;
2215 }
2216 #undef TMPBUFLEN
2217
2218 static int proc_put_char(void __user **buf, size_t *size, char c)
2219 {
2220         if (*size) {
2221                 char __user **buffer = (char __user **)buf;
2222                 if (put_user(c, *buffer))
2223                         return -EFAULT;
2224                 (*size)--, (*buffer)++;
2225                 *buf = *buffer;
2226         }
2227         return 0;
2228 }
2229
2230 static int do_proc_dointvec_conv(bool *negp, unsigned long *lvalp,
2231                                  int *valp,
2232                                  int write, void *data)
2233 {
2234         if (write) {
2235                 *valp = *negp ? -*lvalp : *lvalp;
2236         } else {
2237                 int val = *valp;
2238                 if (val < 0) {
2239                         *negp = true;
2240                         *lvalp = (unsigned long)-val;
2241                 } else {
2242                         *negp = false;
2243                         *lvalp = (unsigned long)val;
2244                 }
2245         }
2246         return 0;
2247 }
2248
2249 static const char proc_wspace_sep[] = { ' ', '\t', '\n' };
2250
2251 static int __do_proc_dointvec(void *tbl_data, struct ctl_table *table,
2252                   int write, void __user *buffer,
2253                   size_t *lenp, loff_t *ppos,
2254                   int (*conv)(bool *negp, unsigned long *lvalp, int *valp,
2255                               int write, void *data),
2256                   void *data)
2257 {
2258         int *i, vleft, first = 1, err = 0;
2259         unsigned long page = 0;
2260         size_t left;
2261         char *kbuf;
2262
2263         if (!tbl_data || !table->maxlen || !*lenp || (*ppos && !write)) {
2264                 *lenp = 0;
2265                 return 0;
2266         }
2267
2268         i = (int *) tbl_data;
2269         vleft = table->maxlen / sizeof(*i);
2270         left = *lenp;
2271
2272         if (!conv)
2273                 conv = do_proc_dointvec_conv;
2274
2275         if (write) {
2276                 if (left > PAGE_SIZE - 1)
2277                         left = PAGE_SIZE - 1;
2278                 page = __get_free_page(GFP_TEMPORARY);
2279                 kbuf = (char *) page;
2280                 if (!kbuf)
2281                         return -ENOMEM;
2282                 if (copy_from_user(kbuf, buffer, left)) {
2283                         err = -EFAULT;
2284                         goto free;
2285                 }
2286                 kbuf[left] = 0;
2287         }
2288
2289         for (; left && vleft--; i++, first=0) {
2290                 unsigned long lval;
2291                 bool neg;
2292
2293                 if (write) {
2294                         left -= proc_skip_spaces(&kbuf);
2295
2296                         if (!left)
2297                                 break;
2298                         err = proc_get_long(&kbuf, &left, &lval, &neg,
2299                                              proc_wspace_sep,
2300                                              sizeof(proc_wspace_sep), NULL);
2301                         if (err)
2302                                 break;
2303                         if (conv(&neg, &lval, i, 1, data)) {
2304                                 err = -EINVAL;
2305                                 break;
2306                         }
2307                 } else {
2308                         if (conv(&neg, &lval, i, 0, data)) {
2309                                 err = -EINVAL;
2310                                 break;
2311                         }
2312                         if (!first)
2313                                 err = proc_put_char(&buffer, &left, '\t');
2314                         if (err)
2315                                 break;
2316                         err = proc_put_long(&buffer, &left, lval, neg);
2317                         if (err)
2318                                 break;
2319                 }
2320         }
2321
2322         if (!write && !first && left && !err)
2323                 err = proc_put_char(&buffer, &left, '\n');
2324         if (write && !err && left)
2325                 left -= proc_skip_spaces(&kbuf);
2326 free:
2327         if (write) {
2328                 free_page(page);
2329                 if (first)
2330                         return err ? : -EINVAL;
2331         }
2332         *lenp -= left;
2333         *ppos += *lenp;
2334         return err;
2335 }
2336
2337 static int do_proc_dointvec(struct ctl_table *table, int write,
2338                   void __user *buffer, size_t *lenp, loff_t *ppos,
2339                   int (*conv)(bool *negp, unsigned long *lvalp, int *valp,
2340                               int write, void *data),
2341                   void *data)
2342 {
2343         return __do_proc_dointvec(table->data, table, write,
2344                         buffer, lenp, ppos, conv, data);
2345 }
2346
2347 /**
2348  * proc_dointvec - read a vector of integers
2349  * @table: the sysctl table
2350  * @write: %TRUE if this is a write to the sysctl file
2351  * @buffer: the user buffer
2352  * @lenp: the size of the user buffer
2353  * @ppos: file position
2354  *
2355  * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2356  * values from/to the user buffer, treated as an ASCII string.
2357  *
2358  * Returns 0 on success.
2359  */
2360 int proc_dointvec(struct ctl_table *table, int write,
2361                      void __user *buffer, size_t *lenp, loff_t *ppos)
2362 {
2363     return do_proc_dointvec(table,write,buffer,lenp,ppos,
2364                             NULL,NULL);
2365 }
2366
2367 /*
2368  * Taint values can only be increased
2369  * This means we can safely use a temporary.
2370  */
2371 static int proc_taint(struct ctl_table *table, int write,
2372                                void __user *buffer, size_t *lenp, loff_t *ppos)
2373 {
2374         struct ctl_table t;
2375         unsigned long tmptaint = get_taint();
2376         int err;
2377
2378         if (write && !capable(CAP_SYS_ADMIN))
2379                 return -EPERM;
2380
2381         t = *table;
2382         t.data = &tmptaint;
2383         err = proc_doulongvec_minmax(&t, write, buffer, lenp, ppos);
2384         if (err < 0)
2385                 return err;
2386
2387         if (write) {
2388                 /*
2389                  * Poor man's atomic or. Not worth adding a primitive
2390                  * to everyone's atomic.h for this
2391                  */
2392                 int i;
2393                 for (i = 0; i < BITS_PER_LONG && tmptaint >> i; i++) {
2394                         if ((tmptaint >> i) & 1)
2395                                 add_taint(i);
2396                 }
2397         }
2398
2399         return err;
2400 }
2401
2402 #ifdef CONFIG_PRINTK
2403 static int proc_dmesg_restrict(struct ctl_table *table, int write,
2404                                 void __user *buffer, size_t *lenp, loff_t *ppos)
2405 {
2406         if (write && !capable(CAP_SYS_ADMIN))
2407                 return -EPERM;
2408
2409         return proc_dointvec_minmax(table, write, buffer, lenp, ppos);
2410 }
2411 #endif
2412
2413 struct do_proc_dointvec_minmax_conv_param {
2414         int *min;
2415         int *max;
2416 };
2417
2418 static int do_proc_dointvec_minmax_conv(bool *negp, unsigned long *lvalp,
2419                                         int *valp,
2420                                         int write, void *data)
2421 {
2422         struct do_proc_dointvec_minmax_conv_param *param = data;
2423         if (write) {
2424                 int val = *negp ? -*lvalp : *lvalp;
2425                 if ((param->min && *param->min > val) ||
2426                     (param->max && *param->max < val))
2427                         return -EINVAL;
2428                 *valp = val;
2429         } else {
2430                 int val = *valp;
2431                 if (val < 0) {
2432                         *negp = true;
2433                         *lvalp = (unsigned long)-val;
2434                 } else {
2435                         *negp = false;
2436                         *lvalp = (unsigned long)val;
2437                 }
2438         }
2439         return 0;
2440 }
2441
2442 /**
2443  * proc_dointvec_minmax - read a vector of integers with min/max values
2444  * @table: the sysctl table
2445  * @write: %TRUE if this is a write to the sysctl file
2446  * @buffer: the user buffer
2447  * @lenp: the size of the user buffer
2448  * @ppos: file position
2449  *
2450  * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2451  * values from/to the user buffer, treated as an ASCII string.
2452  *
2453  * This routine will ensure the values are within the range specified by
2454  * table->extra1 (min) and table->extra2 (max).
2455  *
2456  * Returns 0 on success.
2457  */
2458 int proc_dointvec_minmax(struct ctl_table *table, int write,
2459                   void __user *buffer, size_t *lenp, loff_t *ppos)
2460 {
2461         struct do_proc_dointvec_minmax_conv_param param = {
2462                 .min = (int *) table->extra1,
2463                 .max = (int *) table->extra2,
2464         };
2465         return do_proc_dointvec(table, write, buffer, lenp, ppos,
2466                                 do_proc_dointvec_minmax_conv, &param);
2467 }
2468
2469 static int __do_proc_doulongvec_minmax(void *data, struct ctl_table *table, int write,
2470                                      void __user *buffer,
2471                                      size_t *lenp, loff_t *ppos,
2472                                      unsigned long convmul,
2473                                      unsigned long convdiv)
2474 {
2475         unsigned long *i, *min, *max;
2476         int vleft, first = 1, err = 0;
2477         unsigned long page = 0;
2478         size_t left;
2479         char *kbuf;
2480
2481         if (!data || !table->maxlen || !*lenp || (*ppos && !write)) {
2482                 *lenp = 0;
2483                 return 0;
2484         }
2485
2486         i = (unsigned long *) data;
2487         min = (unsigned long *) table->extra1;
2488         max = (unsigned long *) table->extra2;
2489         vleft = table->maxlen / sizeof(unsigned long);
2490         left = *lenp;
2491
2492         if (write) {
2493                 if (left > PAGE_SIZE - 1)
2494                         left = PAGE_SIZE - 1;
2495                 page = __get_free_page(GFP_TEMPORARY);
2496                 kbuf = (char *) page;
2497                 if (!kbuf)
2498                         return -ENOMEM;
2499                 if (copy_from_user(kbuf, buffer, left)) {
2500                         err = -EFAULT;
2501                         goto free;
2502                 }
2503                 kbuf[left] = 0;
2504         }
2505
2506         for (; left && vleft--; i++, first = 0) {
2507                 unsigned long val;
2508
2509                 if (write) {
2510                         bool neg;
2511
2512                         left -= proc_skip_spaces(&kbuf);
2513
2514                         err = proc_get_long(&kbuf, &left, &val, &neg,
2515                                              proc_wspace_sep,
2516                                              sizeof(proc_wspace_sep), NULL);
2517                         if (err)
2518                                 break;
2519                         if (neg)
2520                                 continue;
2521                         if ((min && val < *min) || (max && val > *max))
2522                                 continue;
2523                         *i = val;
2524                 } else {
2525                         val = convdiv * (*i) / convmul;
2526                         if (!first)
2527                                 err = proc_put_char(&buffer, &left, '\t');
2528                         err = proc_put_long(&buffer, &left, val, false);
2529                         if (err)
2530                                 break;
2531                 }
2532         }
2533
2534         if (!write && !first && left && !err)
2535                 err = proc_put_char(&buffer, &left, '\n');
2536         if (write && !err)
2537                 left -= proc_skip_spaces(&kbuf);
2538 free:
2539         if (write) {
2540                 free_page(page);
2541                 if (first)
2542                         return err ? : -EINVAL;
2543         }
2544         *lenp -= left;
2545         *ppos += *lenp;
2546         return err;
2547 }
2548
2549 static int do_proc_doulongvec_minmax(struct ctl_table *table, int write,
2550                                      void __user *buffer,
2551                                      size_t *lenp, loff_t *ppos,
2552                                      unsigned long convmul,
2553                                      unsigned long convdiv)
2554 {
2555         return __do_proc_doulongvec_minmax(table->data, table, write,
2556                         buffer, lenp, ppos, convmul, convdiv);
2557 }
2558
2559 /**
2560  * proc_doulongvec_minmax - read a vector of long integers with min/max values
2561  * @table: the sysctl table
2562  * @write: %TRUE if this is a write to the sysctl file
2563  * @buffer: the user buffer
2564  * @lenp: the size of the user buffer
2565  * @ppos: file position
2566  *
2567  * Reads/writes up to table->maxlen/sizeof(unsigned long) unsigned long
2568  * values from/to the user buffer, treated as an ASCII string.
2569  *
2570  * This routine will ensure the values are within the range specified by
2571  * table->extra1 (min) and table->extra2 (max).
2572  *
2573  * Returns 0 on success.
2574  */
2575 int proc_doulongvec_minmax(struct ctl_table *table, int write,
2576                            void __user *buffer, size_t *lenp, loff_t *ppos)
2577 {
2578     return do_proc_doulongvec_minmax(table, write, buffer, lenp, ppos, 1l, 1l);
2579 }
2580
2581 /**
2582  * proc_doulongvec_ms_jiffies_minmax - read a vector of millisecond values with min/max values
2583  * @table: the sysctl table
2584  * @write: %TRUE if this is a write to the sysctl file
2585  * @buffer: the user buffer
2586  * @lenp: the size of the user buffer
2587  * @ppos: file position
2588  *
2589  * Reads/writes up to table->maxlen/sizeof(unsigned long) unsigned long
2590  * values from/to the user buffer, treated as an ASCII string. The values
2591  * are treated as milliseconds, and converted to jiffies when they are stored.
2592  *
2593  * This routine will ensure the values are within the range specified by
2594  * table->extra1 (min) and table->extra2 (max).
2595  *
2596  * Returns 0 on success.
2597  */
2598 int proc_doulongvec_ms_jiffies_minmax(struct ctl_table *table, int write,
2599                                       void __user *buffer,
2600                                       size_t *lenp, loff_t *ppos)
2601 {
2602     return do_proc_doulongvec_minmax(table, write, buffer,
2603                                      lenp, ppos, HZ, 1000l);
2604 }
2605
2606
2607 static int do_proc_dointvec_jiffies_conv(bool *negp, unsigned long *lvalp,
2608                                          int *valp,
2609                                          int write, void *data)
2610 {
2611         if (write) {
2612                 if (*lvalp > LONG_MAX / HZ)
2613                         return 1;
2614                 *valp = *negp ? -(*lvalp*HZ) : (*lvalp*HZ);
2615         } else {
2616                 int val = *valp;
2617                 unsigned long lval;
2618                 if (val < 0) {
2619                         *negp = true;
2620                         lval = (unsigned long)-val;
2621                 } else {
2622                         *negp = false;
2623                         lval = (unsigned long)val;
2624                 }
2625                 *lvalp = lval / HZ;
2626         }
2627         return 0;
2628 }
2629
2630 static int do_proc_dointvec_userhz_jiffies_conv(bool *negp, unsigned long *lvalp,
2631                                                 int *valp,
2632                                                 int write, void *data)
2633 {
2634         if (write) {
2635                 if (USER_HZ < HZ && *lvalp > (LONG_MAX / HZ) * USER_HZ)
2636                         return 1;
2637                 *valp = clock_t_to_jiffies(*negp ? -*lvalp : *lvalp);
2638         } else {
2639                 int val = *valp;
2640                 unsigned long lval;
2641                 if (val < 0) {
2642                         *negp = true;
2643                         lval = (unsigned long)-val;
2644                 } else {
2645                         *negp = false;
2646                         lval = (unsigned long)val;
2647                 }
2648                 *lvalp = jiffies_to_clock_t(lval);
2649         }
2650         return 0;
2651 }
2652
2653 static int do_proc_dointvec_ms_jiffies_conv(bool *negp, unsigned long *lvalp,
2654                                             int *valp,
2655                                             int write, void *data)
2656 {
2657         if (write) {
2658                 *valp = msecs_to_jiffies(*negp ? -*lvalp : *lvalp);
2659         } else {
2660                 int val = *valp;
2661                 unsigned long lval;
2662                 if (val < 0) {
2663                         *negp = true;
2664                         lval = (unsigned long)-val;
2665                 } else {
2666                         *negp = false;
2667                         lval = (unsigned long)val;
2668                 }
2669                 *lvalp = jiffies_to_msecs(lval);
2670         }
2671         return 0;
2672 }
2673
2674 /**
2675  * proc_dointvec_jiffies - read a vector of integers as seconds
2676  * @table: the sysctl table
2677  * @write: %TRUE if this is a write to the sysctl file
2678  * @buffer: the user buffer
2679  * @lenp: the size of the user buffer
2680  * @ppos: file position
2681  *
2682  * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2683  * values from/to the user buffer, treated as an ASCII string.
2684  * The values read are assumed to be in seconds, and are converted into
2685  * jiffies.
2686  *
2687  * Returns 0 on success.
2688  */
2689 int proc_dointvec_jiffies(struct ctl_table *table, int write,
2690                           void __user *buffer, size_t *lenp, loff_t *ppos)
2691 {
2692     return do_proc_dointvec(table,write,buffer,lenp,ppos,
2693                             do_proc_dointvec_jiffies_conv,NULL);
2694 }
2695
2696 /**
2697  * proc_dointvec_userhz_jiffies - read a vector of integers as 1/USER_HZ seconds
2698  * @table: the sysctl table
2699  * @write: %TRUE if this is a write to the sysctl file
2700  * @buffer: the user buffer
2701  * @lenp: the size of the user buffer
2702  * @ppos: pointer to the file position
2703  *
2704  * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2705  * values from/to the user buffer, treated as an ASCII string.
2706  * The values read are assumed to be in 1/USER_HZ seconds, and
2707  * are converted into jiffies.
2708  *
2709  * Returns 0 on success.
2710  */
2711 int proc_dointvec_userhz_jiffies(struct ctl_table *table, int write,
2712                                  void __user *buffer, size_t *lenp, loff_t *ppos)
2713 {
2714     return do_proc_dointvec(table,write,buffer,lenp,ppos,
2715                             do_proc_dointvec_userhz_jiffies_conv,NULL);
2716 }
2717
2718 /**
2719  * proc_dointvec_ms_jiffies - read a vector of integers as 1 milliseconds
2720  * @table: the sysctl table
2721  * @write: %TRUE if this is a write to the sysctl file
2722  * @buffer: the user buffer
2723  * @lenp: the size of the user buffer
2724  * @ppos: file position
2725  * @ppos: the current position in the file
2726  *
2727  * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2728  * values from/to the user buffer, treated as an ASCII string.
2729  * The values read are assumed to be in 1/1000 seconds, and
2730  * are converted into jiffies.
2731  *
2732  * Returns 0 on success.
2733  */
2734 int proc_dointvec_ms_jiffies(struct ctl_table *table, int write,
2735                              void __user *buffer, size_t *lenp, loff_t *ppos)
2736 {
2737         return do_proc_dointvec(table, write, buffer, lenp, ppos,
2738                                 do_proc_dointvec_ms_jiffies_conv, NULL);
2739 }
2740
2741 static int proc_do_cad_pid(struct ctl_table *table, int write,
2742                            void __user *buffer, size_t *lenp, loff_t *ppos)
2743 {
2744         struct pid *new_pid;
2745         pid_t tmp;
2746         int r;
2747
2748         tmp = pid_vnr(cad_pid);
2749
2750         r = __do_proc_dointvec(&tmp, table, write, buffer,
2751                                lenp, ppos, NULL, NULL);
2752         if (r || !write)
2753                 return r;
2754
2755         new_pid = find_get_pid(tmp);
2756         if (!new_pid)
2757                 return -ESRCH;
2758
2759         put_pid(xchg(&cad_pid, new_pid));
2760         return 0;
2761 }
2762
2763 /**
2764  * proc_do_large_bitmap - read/write from/to a large bitmap
2765  * @table: the sysctl table
2766  * @write: %TRUE if this is a write to the sysctl file
2767  * @buffer: the user buffer
2768  * @lenp: the size of the user buffer
2769  * @ppos: file position
2770  *
2771  * The bitmap is stored at table->data and the bitmap length (in bits)
2772  * in table->maxlen.
2773  *
2774  * We use a range comma separated format (e.g. 1,3-4,10-10) so that
2775  * large bitmaps may be represented in a compact manner. Writing into
2776  * the file will clear the bitmap then update it with the given input.
2777  *
2778  * Returns 0 on success.
2779  */
2780 int proc_do_large_bitmap(struct ctl_table *table, int write,
2781                          void __user *buffer, size_t *lenp, loff_t *ppos)
2782 {
2783         int err = 0;
2784         bool first = 1;
2785         size_t left = *lenp;
2786         unsigned long bitmap_len = table->maxlen;
2787         unsigned long *bitmap = (unsigned long *) table->data;
2788         unsigned long *tmp_bitmap = NULL;
2789         char tr_a[] = { '-', ',', '\n' }, tr_b[] = { ',', '\n', 0 }, c;
2790
2791         if (!bitmap_len || !left || (*ppos && !write)) {
2792                 *lenp = 0;
2793                 return 0;
2794         }
2795
2796         if (write) {
2797                 unsigned long page = 0;
2798                 char *kbuf;
2799
2800                 if (left > PAGE_SIZE - 1)
2801                         left = PAGE_SIZE - 1;
2802
2803                 page = __get_free_page(GFP_TEMPORARY);
2804                 kbuf = (char *) page;
2805                 if (!kbuf)
2806                         return -ENOMEM;
2807                 if (copy_from_user(kbuf, buffer, left)) {
2808                         free_page(page);
2809                         return -EFAULT;
2810                 }
2811                 kbuf[left] = 0;
2812
2813                 tmp_bitmap = kzalloc(BITS_TO_LONGS(bitmap_len) * sizeof(unsigned long),
2814                                      GFP_KERNEL);
2815                 if (!tmp_bitmap) {
2816                         free_page(page);
2817                         return -ENOMEM;
2818                 }
2819                 proc_skip_char(&kbuf, &left, '\n');
2820                 while (!err && left) {
2821                         unsigned long val_a, val_b;
2822                         bool neg;
2823
2824                         err = proc_get_long(&kbuf, &left, &val_a, &neg, tr_a,
2825                                              sizeof(tr_a), &c);
2826                         if (err)
2827                                 break;
2828                         if (val_a >= bitmap_len || neg) {
2829                                 err = -EINVAL;
2830                                 break;
2831                         }
2832
2833                         val_b = val_a;
2834                         if (left) {
2835                                 kbuf++;
2836                                 left--;
2837                         }
2838
2839                         if (c == '-') {
2840                                 err = proc_get_long(&kbuf, &left, &val_b,
2841                                                      &neg, tr_b, sizeof(tr_b),
2842                                                      &c);
2843                                 if (err)
2844                                         break;
2845                                 if (val_b >= bitmap_len || neg ||
2846                                     val_a > val_b) {
2847                                         err = -EINVAL;
2848                                         break;
2849                                 }
2850                                 if (left) {
2851                                         kbuf++;
2852                                         left--;
2853                                 }
2854                         }
2855
2856                         while (val_a <= val_b)
2857                                 set_bit(val_a++, tmp_bitmap);
2858
2859                         first = 0;
2860                         proc_skip_char(&kbuf, &left, '\n');
2861                 }
2862                 free_page(page);
2863         } else {
2864                 unsigned long bit_a, bit_b = 0;
2865
2866                 while (left) {
2867                         bit_a = find_next_bit(bitmap, bitmap_len, bit_b);
2868                         if (bit_a >= bitmap_len)
2869                                 break;
2870                         bit_b = find_next_zero_bit(bitmap, bitmap_len,
2871                                                    bit_a + 1) - 1;
2872
2873                         if (!first) {
2874                                 err = proc_put_char(&buffer, &left, ',');
2875                                 if (err)
2876                                         break;
2877                         }
2878                         err = proc_put_long(&buffer, &left, bit_a, false);
2879                         if (err)
2880                                 break;
2881                         if (bit_a != bit_b) {
2882                                 err = proc_put_char(&buffer, &left, '-');
2883                                 if (err)
2884                                         break;
2885                                 err = proc_put_long(&buffer, &left, bit_b, false);
2886                                 if (err)
2887                                         break;
2888                         }
2889
2890                         first = 0; bit_b++;
2891                 }
2892                 if (!err)
2893                         err = proc_put_char(&buffer, &left, '\n');
2894         }
2895
2896         if (!err) {
2897                 if (write) {
2898                         if (*ppos)
2899                                 bitmap_or(bitmap, bitmap, tmp_bitmap, bitmap_len);
2900                         else
2901                                 memcpy(bitmap, tmp_bitmap,
2902                                         BITS_TO_LONGS(bitmap_len) * sizeof(unsigned long));
2903                 }
2904                 kfree(tmp_bitmap);
2905                 *lenp -= left;
2906                 *ppos += *lenp;
2907                 return 0;
2908         } else {
2909                 kfree(tmp_bitmap);
2910                 return err;
2911         }
2912 }
2913
2914 #else /* CONFIG_PROC_SYSCTL */
2915
2916 int proc_dostring(struct ctl_table *table, int write,
2917                   void __user *buffer, size_t *lenp, loff_t *ppos)
2918 {
2919         return -ENOSYS;
2920 }
2921
2922 int proc_dointvec(struct ctl_table *table, int write,
2923                   void __user *buffer, size_t *lenp, loff_t *ppos)
2924 {
2925         return -ENOSYS;
2926 }
2927
2928 int proc_dointvec_minmax(struct ctl_table *table, int write,
2929                     void __user *buffer, size_t *lenp, loff_t *ppos)
2930 {
2931         return -ENOSYS;
2932 }
2933
2934 int proc_dointvec_jiffies(struct ctl_table *table, int write,
2935                     void __user *buffer, size_t *lenp, loff_t *ppos)
2936 {
2937         return -ENOSYS;
2938 }
2939
2940 int proc_dointvec_userhz_jiffies(struct ctl_table *table, int write,
2941                     void __user *buffer, size_t *lenp, loff_t *ppos)
2942 {
2943         return -ENOSYS;
2944 }
2945
2946 int proc_dointvec_ms_jiffies(struct ctl_table *table, int write,
2947                              void __user *buffer, size_t *lenp, loff_t *ppos)
2948 {
2949         return -ENOSYS;
2950 }
2951
2952 int proc_doulongvec_minmax(struct ctl_table *table, int write,
2953                     void __user *buffer, size_t *lenp, loff_t *ppos)
2954 {
2955         return -ENOSYS;
2956 }
2957
2958 int proc_doulongvec_ms_jiffies_minmax(struct ctl_table *table, int write,
2959                                       void __user *buffer,
2960                                       size_t *lenp, loff_t *ppos)
2961 {
2962     return -ENOSYS;
2963 }
2964
2965
2966 #endif /* CONFIG_PROC_SYSCTL */
2967
2968 /*
2969  * No sense putting this after each symbol definition, twice,
2970  * exception granted :-)
2971  */
2972 EXPORT_SYMBOL(proc_dointvec);
2973 EXPORT_SYMBOL(proc_dointvec_jiffies);
2974 EXPORT_SYMBOL(proc_dointvec_minmax);
2975 EXPORT_SYMBOL(proc_dointvec_userhz_jiffies);
2976 EXPORT_SYMBOL(proc_dointvec_ms_jiffies);
2977 EXPORT_SYMBOL(proc_dostring);
2978 EXPORT_SYMBOL(proc_doulongvec_minmax);
2979 EXPORT_SYMBOL(proc_doulongvec_ms_jiffies_minmax);
2980 EXPORT_SYMBOL(register_sysctl_table);
2981 EXPORT_SYMBOL(register_sysctl_paths);
2982 EXPORT_SYMBOL(unregister_sysctl_table);