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,
 734                 .extra1         = &zero,
 735                 .extra2         = &one,
 736         },
 737         {
 738                 .procname       = "watchdog_thresh",
 739                 .data           = &watchdog_thresh,
 740                 .maxlen         = sizeof(int),
 741                 .mode           = 0644,
 742                 .proc_handler   = proc_dowatchdog,
 743                 .extra1         = &neg_one,
 744                 .extra2         = &sixty,
 745         },
 746         {
 747                 .procname       = "softlockup_panic",
 748                 .data           = &softlockup_panic,
 749                 .maxlen         = sizeof(int),
 750                 .mode           = 0644,
 751                 .proc_handler   = proc_dointvec_minmax,
 752                 .extra1         = &zero,
 753                 .extra2         = &one,
 754         },
 755         {
 756                 .procname       = "nmi_watchdog",
 757                 .data           = &watchdog_enabled,
 758                 .maxlen         = sizeof (int),
 759                 .mode           = 0644,
 760                 .proc_handler   = proc_dowatchdog,
 761                 .extra1         = &zero,
 762                 .extra2         = &one,
 763         },
 764 #endif
 765 #if defined(CONFIG_X86_LOCAL_APIC) && defined(CONFIG_X86)
 766         {
 767                 .procname       = "unknown_nmi_panic",
 768                 .data           = &unknown_nmi_panic,
 769                 .maxlen         = sizeof (int),
 770                 .mode           = 0644,
 771                 .proc_handler   = proc_dointvec,
 772         },
 773 #endif
 774 #if defined(CONFIG_X86)
 775         {
 776                 .procname       = "panic_on_unrecovered_nmi",
 777                 .data           = &panic_on_unrecovered_nmi,
 778                 .maxlen         = sizeof(int),
 779                 .mode           = 0644,
 780                 .proc_handler   = proc_dointvec,
 781         },
 782         {
 783                 .procname       = "panic_on_io_nmi",
 784                 .data           = &panic_on_io_nmi,
 785                 .maxlen         = sizeof(int),
 786                 .mode           = 0644,
 787                 .proc_handler   = proc_dointvec,
 788         },
 789         {
 790                 .procname       = "bootloader_type",
 791                 .data           = &bootloader_type,
 792                 .maxlen         = sizeof (int),
 793                 .mode           = 0444,
 794                 .proc_handler   = proc_dointvec,
 795         },
 796         {
 797                 .procname       = "bootloader_version",
 798                 .data           = &bootloader_version,
 799                 .maxlen         = sizeof (int),
 800                 .mode           = 0444,
 801                 .proc_handler   = proc_dointvec,
 802         },
 803         {
 804                 .procname       = "kstack_depth_to_print",
 805                 .data           = &kstack_depth_to_print,
 806                 .maxlen         = sizeof(int),
 807                 .mode           = 0644,
 808                 .proc_handler   = proc_dointvec,
 809         },
 810         {
 811                 .procname       = "io_delay_type",
 812                 .data           = &io_delay_type,
 813                 .maxlen         = sizeof(int),
 814                 .mode           = 0644,
 815                 .proc_handler   = proc_dointvec,
 816         },
 817 #endif
 818 #if defined(CONFIG_MMU)
 819         {
 820                 .procname       = "randomize_va_space",
 821                 .data           = &randomize_va_space,
 822                 .maxlen         = sizeof(int),
 823                 .mode           = 0644,
 824                 .proc_handler   = proc_dointvec,
 825         },
 826 #endif
 827 #if defined(CONFIG_S390) && defined(CONFIG_SMP)
 828         {
 829                 .procname       = "spin_retry",
 830                 .data           = &spin_retry,
 831                 .maxlen         = sizeof (int),
 832                 .mode           = 0644,
 833                 .proc_handler   = proc_dointvec,
 834         },
 835 #endif
 836 #if     defined(CONFIG_ACPI_SLEEP) && defined(CONFIG_X86)
 837         {
 838                 .procname       = "acpi_video_flags",
 839                 .data           = &acpi_realmode_flags,
 840                 .maxlen         = sizeof (unsigned long),
 841                 .mode           = 0644,
 842                 .proc_handler   = proc_doulongvec_minmax,
 843         },
 844 #endif
 845 #ifdef CONFIG_IA64
 846         {
 847                 .procname       = "ignore-unaligned-usertrap",
 848                 .data           = &no_unaligned_warning,
 849                 .maxlen         = sizeof (int),
 850                 .mode           = 0644,
 851                 .proc_handler   = proc_dointvec,
 852         },
 853         {
 854                 .procname       = "unaligned-dump-stack",
 855                 .data           = &unaligned_dump_stack,
 856                 .maxlen         = sizeof (int),
 857                 .mode           = 0644,
 858                 .proc_handler   = proc_dointvec,
 859         },
 860 #endif
 861 #ifdef CONFIG_DETECT_HUNG_TASK
 862         {
 863                 .procname       = "hung_task_panic",
 864                 .data           = &sysctl_hung_task_panic,
 865                 .maxlen         = sizeof(int),
 866                 .mode           = 0644,
 867                 .proc_handler   = proc_dointvec_minmax,
 868                 .extra1         = &zero,
 869                 .extra2         = &one,
 870         },
 871         {
 872                 .procname       = "hung_task_check_count",
 873                 .data           = &sysctl_hung_task_check_count,
 874                 .maxlen         = sizeof(unsigned long),
 875                 .mode           = 0644,
 876                 .proc_handler   = proc_doulongvec_minmax,
 877         },
 878         {
 879                 .procname       = "hung_task_timeout_secs",
 880                 .data           = &sysctl_hung_task_timeout_secs,
 881                 .maxlen         = sizeof(unsigned long),
 882                 .mode           = 0644,
 883                 .proc_handler   = proc_dohung_task_timeout_secs,
 884         },
 885         {
 886                 .procname       = "hung_task_warnings",
 887                 .data           = &sysctl_hung_task_warnings,
 888                 .maxlen         = sizeof(unsigned long),
 889                 .mode           = 0644,
 890                 .proc_handler   = proc_doulongvec_minmax,
 891         },
 892 #endif
 893 #ifdef CONFIG_COMPAT
 894         {
 895                 .procname       = "compat-log",
 896                 .data           = &compat_log,
 897                 .maxlen         = sizeof (int),
 898                 .mode           = 0644,
 899                 .proc_handler   = proc_dointvec,
 900         },
 901 #endif
 902 #ifdef CONFIG_RT_MUTEXES
 903         {
 904                 .procname       = "max_lock_depth",
 905                 .data           = &max_lock_depth,
 906                 .maxlen         = sizeof(int),
 907                 .mode           = 0644,
 908                 .proc_handler   = proc_dointvec,
 909         },
 910 #endif
 911         {
 912                 .procname       = "poweroff_cmd",
 913                 .data           = &poweroff_cmd,
 914                 .maxlen         = POWEROFF_CMD_PATH_LEN,
 915                 .mode           = 0644,
 916                 .proc_handler   = proc_dostring,
 917         },
 918 #ifdef CONFIG_KEYS
 919         {
 920                 .procname       = "keys",
 921                 .mode           = 0555,
 922                 .child          = key_sysctls,
 923         },
 924 #endif
 925 #ifdef CONFIG_RCU_TORTURE_TEST
 926         {
 927                 .procname       = "rcutorture_runnable",
 928                 .data           = &rcutorture_runnable,
 929                 .maxlen         = sizeof(int),
 930                 .mode           = 0644,
 931                 .proc_handler   = proc_dointvec,
 932         },
 933 #endif
 934 #ifdef CONFIG_PERF_EVENTS
 935         {
 936                 .procname       = "perf_event_paranoid",
 937                 .data           = &sysctl_perf_event_paranoid,
 938                 .maxlen         = sizeof(sysctl_perf_event_paranoid),
 939                 .mode           = 0644,
 940                 .proc_handler   = proc_dointvec,
 941         },
 942         {
 943                 .procname       = "perf_event_mlock_kb",
 944                 .data           = &sysctl_perf_event_mlock,
 945                 .maxlen         = sizeof(sysctl_perf_event_mlock),
 946                 .mode           = 0644,
 947                 .proc_handler   = proc_dointvec,
 948         },
 949         {
 950                 .procname       = "perf_event_max_sample_rate",
 951                 .data           = &sysctl_perf_event_sample_rate,
 952                 .maxlen         = sizeof(sysctl_perf_event_sample_rate),
 953                 .mode           = 0644,
 954                 .proc_handler   = perf_proc_update_handler,
 955         },
 956 #endif
 957 #ifdef CONFIG_KMEMCHECK
 958         {
 959                 .procname       = "kmemcheck",
 960                 .data           = &kmemcheck_enabled,
 961                 .maxlen         = sizeof(int),
 962                 .mode           = 0644,
 963                 .proc_handler   = proc_dointvec,
 964         },
 965 #endif
 966 #ifdef CONFIG_BLOCK
 967         {
 968                 .procname       = "blk_iopoll",
 969                 .data           = &blk_iopoll_enabled,
 970                 .maxlen         = sizeof(int),
 971                 .mode           = 0644,
 972                 .proc_handler   = proc_dointvec,
 973         },
 974 #endif
 975         { }
 976 };
 977
 978 static struct ctl_table vm_table[] = {
 979         {
 980                 .procname       = "overcommit_memory",
 981                 .data           = &sysctl_overcommit_memory,
 982                 .maxlen         = sizeof(sysctl_overcommit_memory),
 983                 .mode           = 0644,
 984                 .proc_handler   = proc_dointvec_minmax,
 985                 .extra1         = &zero,
 986                 .extra2         = &two,
 987         },
 988         {
 989                 .procname       = "panic_on_oom",
 990                 .data           = &sysctl_panic_on_oom,
 991                 .maxlen         = sizeof(sysctl_panic_on_oom),
 992                 .mode           = 0644,
 993                 .proc_handler   = proc_dointvec_minmax,
 994                 .extra1         = &zero,
 995                 .extra2         = &two,
 996         },
 997         {
 998                 .procname       = "oom_kill_allocating_task",
 999                 .data           = &sysctl_oom_kill_allocating_task,
1000                 .maxlen         = sizeof(sysctl_oom_kill_allocating_task),
1001                 .mode           = 0644,
1002                 .proc_handler   = proc_dointvec,
1003         },
1004         {
1005                 .procname       = "oom_dump_tasks",
1006                 .data           = &sysctl_oom_dump_tasks,
1007                 .maxlen         = sizeof(sysctl_oom_dump_tasks),
1008                 .mode           = 0644,
1009                 .proc_handler   = proc_dointvec,
1010         },
1011         {
1012                 .procname       = "overcommit_ratio",
1013                 .data           = &sysctl_overcommit_ratio,
1014                 .maxlen         = sizeof(sysctl_overcommit_ratio),
1015                 .mode           = 0644,
1016                 .proc_handler   = proc_dointvec,
1017         },
1018         {
1019                 .procname       = "page-cluster",
1020                 .data           = &page_cluster,
1021                 .maxlen         = sizeof(int),
1022                 .mode           = 0644,
1023                 .proc_handler   = proc_dointvec_minmax,
1024                 .extra1         = &zero,
1025         },
1026         {
1027                 .procname       = "dirty_background_ratio",
1028                 .data           = &dirty_background_ratio,
1029                 .maxlen         = sizeof(dirty_background_ratio),
1030                 .mode           = 0644,
1031                 .proc_handler   = dirty_background_ratio_handler,
1032                 .extra1         = &zero,
1033                 .extra2         = &one_hundred,
1034         },
1035         {
1036                 .procname       = "dirty_background_bytes",
1037                 .data           = &dirty_background_bytes,
1038                 .maxlen         = sizeof(dirty_background_bytes),
1039                 .mode           = 0644,
1040                 .proc_handler   = dirty_background_bytes_handler,
1041                 .extra1         = &one_ul,
1042         },
1043         {
1044                 .procname       = "dirty_ratio",
1045                 .data           = &vm_dirty_ratio,
1046                 .maxlen         = sizeof(vm_dirty_ratio),
1047                 .mode           = 0644,
1048                 .proc_handler   = dirty_ratio_handler,
1049                 .extra1         = &zero,
1050                 .extra2         = &one_hundred,
1051         },
1052         {
1053                 .procname       = "dirty_bytes",
1054                 .data           = &vm_dirty_bytes,
1055                 .maxlen         = sizeof(vm_dirty_bytes),
1056                 .mode           = 0644,
1057                 .proc_handler   = dirty_bytes_handler,
1058                 .extra1         = &dirty_bytes_min,
1059         },
1060         {
1061                 .procname       = "dirty_writeback_centisecs",
1062                 .data           = &dirty_writeback_interval,
1063                 .maxlen         = sizeof(dirty_writeback_interval),
1064                 .mode           = 0644,
1065                 .proc_handler   = dirty_writeback_centisecs_handler,
1066         },
1067         {
1068                 .procname       = "dirty_expire_centisecs",
1069                 .data           = &dirty_expire_interval,
1070                 .maxlen         = sizeof(dirty_expire_interval),
1071                 .mode           = 0644,
1072                 .proc_handler   = proc_dointvec_minmax,
1073                 .extra1         = &zero,
1074         },
1075         {
1076                 .procname       = "nr_pdflush_threads",
1077                 .data           = &nr_pdflush_threads,
1078                 .maxlen         = sizeof nr_pdflush_threads,
1079                 .mode           = 0444 /* read-only*/,
1080                 .proc_handler   = proc_dointvec,
1081         },
1082         {
1083                 .procname       = "swappiness",
1084                 .data           = &vm_swappiness,
1085                 .maxlen         = sizeof(vm_swappiness),
1086                 .mode           = 0644,
1087                 .proc_handler   = proc_dointvec_minmax,
1088                 .extra1         = &zero,
1089                 .extra2         = &one_hundred,
1090         },
1091 #ifdef CONFIG_HUGETLB_PAGE
1092         {
1093                 .procname       = "nr_hugepages",
1094                 .data           = NULL,
1095                 .maxlen         = sizeof(unsigned long),
1096                 .mode           = 0644,
1097                 .proc_handler   = hugetlb_sysctl_handler,
1098                 .extra1         = (void *)&hugetlb_zero,
1099                 .extra2         = (void *)&hugetlb_infinity,
1100         },
1101 #ifdef CONFIG_NUMA
1102         {
1103                 .procname       = "nr_hugepages_mempolicy",
1104                 .data           = NULL,
1105                 .maxlen         = sizeof(unsigned long),
1106                 .mode           = 0644,
1107                 .proc_handler   = &hugetlb_mempolicy_sysctl_handler,
1108                 .extra1         = (void *)&hugetlb_zero,
1109                 .extra2         = (void *)&hugetlb_infinity,
1110         },
1111 #endif
1112          {
1113                 .procname       = "hugetlb_shm_group",
1114                 .data           = &sysctl_hugetlb_shm_group,
1115                 .maxlen         = sizeof(gid_t),
1116                 .mode           = 0644,
1117                 .proc_handler   = proc_dointvec,
1118          },
1119          {
1120                 .procname       = "hugepages_treat_as_movable",
1121                 .data           = &hugepages_treat_as_movable,
1122                 .maxlen         = sizeof(int),
1123                 .mode           = 0644,
1124                 .proc_handler   = hugetlb_treat_movable_handler,
1125         },
1126         {
1127                 .procname       = "nr_overcommit_hugepages",
1128                 .data           = NULL,
1129                 .maxlen         = sizeof(unsigned long),
1130                 .mode           = 0644,
1131                 .proc_handler   = hugetlb_overcommit_handler,
1132                 .extra1         = (void *)&hugetlb_zero,
1133                 .extra2         = (void *)&hugetlb_infinity,
1134         },
1135 #endif
1136         {
1137                 .procname       = "lowmem_reserve_ratio",
1138                 .data           = &sysctl_lowmem_reserve_ratio,
1139                 .maxlen         = sizeof(sysctl_lowmem_reserve_ratio),
1140                 .mode           = 0644,
1141                 .proc_handler   = lowmem_reserve_ratio_sysctl_handler,
1142         },
1143         {
1144                 .procname       = "drop_caches",
1145                 .data           = &sysctl_drop_caches,
1146                 .maxlen         = sizeof(int),
1147                 .mode           = 0644,
1148                 .proc_handler   = drop_caches_sysctl_handler,
1149                 .extra1         = &one,
1150                 .extra2         = &three,
1151         },
1152 #ifdef CONFIG_COMPACTION
1153         {
1154                 .procname       = "compact_memory",
1155                 .data           = &sysctl_compact_memory,
1156                 .maxlen         = sizeof(int),
1157                 .mode           = 0200,
1158                 .proc_handler   = sysctl_compaction_handler,
1159         },
1160         {
1161                 .procname       = "extfrag_threshold",
1162                 .data           = &sysctl_extfrag_threshold,
1163                 .maxlen         = sizeof(int),
1164                 .mode           = 0644,
1165                 .proc_handler   = sysctl_extfrag_handler,
1166                 .extra1         = &min_extfrag_threshold,
1167                 .extra2         = &max_extfrag_threshold,
1168         },
1169
1170 #endif /* CONFIG_COMPACTION */
1171         {
1172                 .procname       = "min_free_kbytes",
1173                 .data           = &min_free_kbytes,
1174                 .maxlen         = sizeof(min_free_kbytes),
1175                 .mode           = 0644,
1176                 .proc_handler   = min_free_kbytes_sysctl_handler,
1177                 .extra1         = &zero,
1178         },
1179         {
1180                 .procname       = "percpu_pagelist_fraction",
1181                 .data           = &percpu_pagelist_fraction,
1182                 .maxlen         = sizeof(percpu_pagelist_fraction),
1183                 .mode           = 0644,
1184                 .proc_handler   = percpu_pagelist_fraction_sysctl_handler,
1185                 .extra1         = &min_percpu_pagelist_fract,
1186         },
1187 #ifdef CONFIG_MMU
1188         {
1189                 .procname       = "max_map_count",
1190                 .data           = &sysctl_max_map_count,
1191                 .maxlen         = sizeof(sysctl_max_map_count),
1192                 .mode           = 0644,
1193                 .proc_handler   = proc_dointvec_minmax,
1194                 .extra1         = &zero,
1195         },
1196 #else
1197         {
1198                 .procname       = "nr_trim_pages",
1199                 .data           = &sysctl_nr_trim_pages,
1200                 .maxlen         = sizeof(sysctl_nr_trim_pages),
1201                 .mode           = 0644,
1202                 .proc_handler   = proc_dointvec_minmax,
1203                 .extra1         = &zero,
1204         },
1205 #endif
1206         {
1207                 .procname       = "laptop_mode",
1208                 .data           = &laptop_mode,
1209                 .maxlen         = sizeof(laptop_mode),
1210                 .mode           = 0644,
1211                 .proc_handler   = proc_dointvec_jiffies,
1212         },
1213         {
1214                 .procname       = "block_dump",
1215                 .data           = &block_dump,
1216                 .maxlen         = sizeof(block_dump),
1217                 .mode           = 0644,
1218                 .proc_handler   = proc_dointvec,
1219                 .extra1         = &zero,
1220         },
1221         {
1222                 .procname       = "vfs_cache_pressure",
1223                 .data           = &sysctl_vfs_cache_pressure,
1224                 .maxlen         = sizeof(sysctl_vfs_cache_pressure),
1225                 .mode           = 0644,
1226                 .proc_handler   = proc_dointvec,
1227                 .extra1         = &zero,
1228         },
1229 #ifdef HAVE_ARCH_PICK_MMAP_LAYOUT
1230         {
1231                 .procname       = "legacy_va_layout",
1232                 .data           = &sysctl_legacy_va_layout,
1233                 .maxlen         = sizeof(sysctl_legacy_va_layout),
1234                 .mode           = 0644,
1235                 .proc_handler   = proc_dointvec,
1236                 .extra1         = &zero,
1237         },
1238 #endif
1239 #ifdef CONFIG_NUMA
1240         {
1241                 .procname       = "zone_reclaim_mode",
1242                 .data           = &zone_reclaim_mode,
1243                 .maxlen         = sizeof(zone_reclaim_mode),
1244                 .mode           = 0644,
1245                 .proc_handler   = proc_dointvec,
1246                 .extra1         = &zero,
1247         },
1248         {
1249                 .procname       = "min_unmapped_ratio",
1250                 .data           = &sysctl_min_unmapped_ratio,
1251                 .maxlen         = sizeof(sysctl_min_unmapped_ratio),
1252                 .mode           = 0644,
1253                 .proc_handler   = sysctl_min_unmapped_ratio_sysctl_handler,
1254                 .extra1         = &zero,
1255                 .extra2         = &one_hundred,
1256         },
1257         {
1258                 .procname       = "min_slab_ratio",
1259                 .data           = &sysctl_min_slab_ratio,
1260                 .maxlen         = sizeof(sysctl_min_slab_ratio),
1261                 .mode           = 0644,
1262                 .proc_handler   = sysctl_min_slab_ratio_sysctl_handler,
1263                 .extra1         = &zero,
1264                 .extra2         = &one_hundred,
1265         },
1266 #endif
1267 #ifdef CONFIG_SMP
1268         {
1269                 .procname       = "stat_interval",
1270                 .data           = &sysctl_stat_interval,
1271                 .maxlen         = sizeof(sysctl_stat_interval),
1272                 .mode           = 0644,
1273                 .proc_handler   = proc_dointvec_jiffies,
1274         },
1275 #endif
1276 #ifdef CONFIG_MMU
1277         {
1278                 .procname       = "mmap_min_addr",
1279                 .data           = &dac_mmap_min_addr,
1280                 .maxlen         = sizeof(unsigned long),
1281                 .mode           = 0644,
1282                 .proc_handler   = mmap_min_addr_handler,
1283         },
1284 #endif
1285 #ifdef CONFIG_NUMA
1286         {
1287                 .procname       = "numa_zonelist_order",
1288                 .data           = &numa_zonelist_order,
1289                 .maxlen         = NUMA_ZONELIST_ORDER_LEN,
1290                 .mode           = 0644,
1291                 .proc_handler   = numa_zonelist_order_handler,
1292         },
1293 #endif
1294 #if (defined(CONFIG_X86_32) && !defined(CONFIG_UML))|| \
1295    (defined(CONFIG_SUPERH) && defined(CONFIG_VSYSCALL))
1296         {
1297                 .procname       = "vdso_enabled",
1298                 .data           = &vdso_enabled,
1299                 .maxlen         = sizeof(vdso_enabled),
1300                 .mode           = 0644,
1301                 .proc_handler   = proc_dointvec,
1302                 .extra1         = &zero,
1303         },
1304 #endif
1305 #ifdef CONFIG_HIGHMEM
1306         {
1307                 .procname       = "highmem_is_dirtyable",
1308                 .data           = &vm_highmem_is_dirtyable,
1309                 .maxlen         = sizeof(vm_highmem_is_dirtyable),
1310                 .mode           = 0644,
1311                 .proc_handler   = proc_dointvec_minmax,
1312                 .extra1         = &zero,
1313                 .extra2         = &one,
1314         },
1315 #endif
1316         {
1317                 .procname       = "scan_unevictable_pages",
1318                 .data           = &scan_unevictable_pages,
1319                 .maxlen         = sizeof(scan_unevictable_pages),
1320                 .mode           = 0644,
1321                 .proc_handler   = scan_unevictable_handler,
1322         },
1323 #ifdef CONFIG_MEMORY_FAILURE
1324         {
1325                 .procname       = "memory_failure_early_kill",
1326                 .data           = &sysctl_memory_failure_early_kill,
1327                 .maxlen         = sizeof(sysctl_memory_failure_early_kill),
1328                 .mode           = 0644,
1329                 .proc_handler   = proc_dointvec_minmax,
1330                 .extra1         = &zero,
1331                 .extra2         = &one,
1332         },
1333         {
1334                 .procname       = "memory_failure_recovery",
1335                 .data           = &sysctl_memory_failure_recovery,
1336                 .maxlen         = sizeof(sysctl_memory_failure_recovery),
1337                 .mode           = 0644,
1338                 .proc_handler   = proc_dointvec_minmax,
1339                 .extra1         = &zero,
1340                 .extra2         = &one,
1341         },
1342 #endif
1343         { }
1344 };
1345
1346 #if defined(CONFIG_BINFMT_MISC) || defined(CONFIG_BINFMT_MISC_MODULE)
1347 static struct ctl_table binfmt_misc_table[] = {
1348         { }
1349 };
1350 #endif
1351
1352 static struct ctl_table fs_table[] = {
1353         {
1354                 .procname       = "inode-nr",
1355                 .data           = &inodes_stat,
1356                 .maxlen         = 2*sizeof(int),
1357                 .mode           = 0444,
1358                 .proc_handler   = proc_nr_inodes,
1359         },
1360         {
1361                 .procname       = "inode-state",
1362                 .data           = &inodes_stat,
1363                 .maxlen         = 7*sizeof(int),
1364                 .mode           = 0444,
1365                 .proc_handler   = proc_nr_inodes,
1366         },
1367         {
1368                 .procname       = "file-nr",
1369                 .data           = &files_stat,
1370                 .maxlen         = sizeof(files_stat),
1371                 .mode           = 0444,
1372                 .proc_handler   = proc_nr_files,
1373         },
1374         {
1375                 .procname       = "file-max",
1376                 .data           = &files_stat.max_files,
1377                 .maxlen         = sizeof(files_stat.max_files),
1378                 .mode           = 0644,
1379                 .proc_handler   = proc_doulongvec_minmax,
1380         },
1381         {
1382                 .procname       = "nr_open",
1383                 .data           = &sysctl_nr_open,
1384                 .maxlen         = sizeof(int),
1385                 .mode           = 0644,
1386                 .proc_handler   = proc_dointvec_minmax,
1387                 .extra1         = &sysctl_nr_open_min,
1388                 .extra2         = &sysctl_nr_open_max,
1389         },
1390         {
1391                 .procname       = "dentry-state",
1392                 .data           = &dentry_stat,
1393                 .maxlen         = 6*sizeof(int),
1394                 .mode           = 0444,
1395                 .proc_handler   = proc_nr_dentry,
1396         },
1397         {
1398                 .procname       = "overflowuid",
1399                 .data           = &fs_overflowuid,
1400                 .maxlen         = sizeof(int),
1401                 .mode           = 0644,
1402                 .proc_handler   = proc_dointvec_minmax,
1403                 .extra1         = &minolduid,
1404                 .extra2         = &maxolduid,
1405         },
1406         {
1407                 .procname       = "overflowgid",
1408                 .data           = &fs_overflowgid,
1409                 .maxlen         = sizeof(int),
1410                 .mode           = 0644,
1411                 .proc_handler   = proc_dointvec_minmax,
1412                 .extra1         = &minolduid,
1413                 .extra2         = &maxolduid,
1414         },
1415 #ifdef CONFIG_FILE_LOCKING
1416         {
1417                 .procname       = "leases-enable",
1418                 .data           = &leases_enable,
1419                 .maxlen         = sizeof(int),
1420                 .mode           = 0644,
1421                 .proc_handler   = proc_dointvec,
1422         },
1423 #endif
1424 #ifdef CONFIG_DNOTIFY
1425         {
1426                 .procname       = "dir-notify-enable",
1427                 .data           = &dir_notify_enable,
1428                 .maxlen         = sizeof(int),
1429                 .mode           = 0644,
1430                 .proc_handler   = proc_dointvec,
1431         },
1432 #endif
1433 #ifdef CONFIG_MMU
1434 #ifdef CONFIG_FILE_LOCKING
1435         {
1436                 .procname       = "lease-break-time",
1437                 .data           = &lease_break_time,
1438                 .maxlen         = sizeof(int),
1439                 .mode           = 0644,
1440                 .proc_handler   = proc_dointvec,
1441         },
1442 #endif
1443 #ifdef CONFIG_AIO
1444         {
1445                 .procname       = "aio-nr",
1446                 .data           = &aio_nr,
1447                 .maxlen         = sizeof(aio_nr),
1448                 .mode           = 0444,
1449                 .proc_handler   = proc_doulongvec_minmax,
1450         },
1451         {
1452                 .procname       = "aio-max-nr",
1453                 .data           = &aio_max_nr,
1454                 .maxlen         = sizeof(aio_max_nr),
1455                 .mode           = 0644,
1456                 .proc_handler   = proc_doulongvec_minmax,
1457         },
1458 #endif /* CONFIG_AIO */
1459 #ifdef CONFIG_INOTIFY_USER
1460         {
1461                 .procname       = "inotify",
1462                 .mode           = 0555,
1463                 .child          = inotify_table,
1464         },
1465 #endif
1466 #ifdef CONFIG_EPOLL
1467         {
1468                 .procname       = "epoll",
1469                 .mode           = 0555,
1470                 .child          = epoll_table,
1471         },
1472 #endif
1473 #endif
1474         {
1475                 .procname       = "suid_dumpable",
1476                 .data           = &suid_dumpable,
1477                 .maxlen         = sizeof(int),
1478                 .mode           = 0644,
1479                 .proc_handler   = proc_dointvec_minmax,
1480                 .extra1         = &zero,
1481                 .extra2         = &two,
1482         },
1483 #if defined(CONFIG_BINFMT_MISC) || defined(CONFIG_BINFMT_MISC_MODULE)
1484         {
1485                 .procname       = "binfmt_misc",
1486                 .mode           = 0555,
1487                 .child          = binfmt_misc_table,
1488         },
1489 #endif
1490         {
1491                 .procname       = "pipe-max-size",
1492                 .data           = &pipe_max_size,
1493                 .maxlen         = sizeof(int),
1494                 .mode           = 0644,
1495                 .proc_handler   = &pipe_proc_fn,
1496                 .extra1         = &pipe_min_size,
1497         },
1498         { }
1499 };
1500
1501 static struct ctl_table debug_table[] = {
1502 #if defined(CONFIG_X86) || defined(CONFIG_PPC) || defined(CONFIG_SPARC) || \
1503     defined(CONFIG_S390)
1504         {
1505                 .procname       = "exception-trace",
1506                 .data           = &show_unhandled_signals,
1507                 .maxlen         = sizeof(int),
1508                 .mode           = 0644,
1509                 .proc_handler   = proc_dointvec
1510         },
1511 #endif
1512 #if defined(CONFIG_OPTPROBES)
1513         {
1514                 .procname       = "kprobes-optimization",
1515                 .data           = &sysctl_kprobes_optimization,
1516                 .maxlen         = sizeof(int),
1517                 .mode           = 0644,
1518                 .proc_handler   = proc_kprobes_optimization_handler,
1519                 .extra1         = &zero,
1520                 .extra2         = &one,
1521         },
1522 #endif
1523         { }
1524 };
1525
1526 static struct ctl_table dev_table[] = {
1527         { }
1528 };
1529
1530 static DEFINE_SPINLOCK(sysctl_lock);
1531
1532 /* called under sysctl_lock */
1533 static int use_table(struct ctl_table_header *p)
1534 {
1535         if (unlikely(p->unregistering))
1536                 return 0;
1537         p->used++;
1538         return 1;
1539 }
1540
1541 /* called under sysctl_lock */
1542 static void unuse_table(struct ctl_table_header *p)
1543 {
1544         if (!--p->used)
1545                 if (unlikely(p->unregistering))
1546                         complete(p->unregistering);
1547 }
1548
1549 /* called under sysctl_lock, will reacquire if has to wait */
1550 static void start_unregistering(struct ctl_table_header *p)
1551 {
1552         /*
1553          * if p->used is 0, nobody will ever touch that entry again;
1554          * we'll eliminate all paths to it before dropping sysctl_lock
1555          */
1556         if (unlikely(p->used)) {
1557                 struct completion wait;
1558                 init_completion(&wait);
1559                 p->unregistering = &wait;
1560                 spin_unlock(&sysctl_lock);
1561                 wait_for_completion(&wait);
1562                 spin_lock(&sysctl_lock);
1563         } else {
1564                 /* anything non-NULL; we'll never dereference it */
1565                 p->unregistering = ERR_PTR(-EINVAL);
1566         }
1567         /*
1568          * do not remove from the list until nobody holds it; walking the
1569          * list in do_sysctl() relies on that.
1570          */
1571         list_del_init(&p->ctl_entry);
1572 }
1573
1574 void sysctl_head_get(struct ctl_table_header *head)
1575 {
1576         spin_lock(&sysctl_lock);
1577         head->count++;
1578         spin_unlock(&sysctl_lock);
1579 }
1580
1581 static void free_head(struct rcu_head *rcu)
1582 {
1583         kfree(container_of(rcu, struct ctl_table_header, rcu));
1584 }
1585
1586 void sysctl_head_put(struct ctl_table_header *head)
1587 {
1588         spin_lock(&sysctl_lock);
1589         if (!--head->count)
1590                 call_rcu(&head->rcu, free_head);
1591         spin_unlock(&sysctl_lock);
1592 }
1593
1594 struct ctl_table_header *sysctl_head_grab(struct ctl_table_header *head)
1595 {
1596         if (!head)
1597                 BUG();
1598         spin_lock(&sysctl_lock);
1599         if (!use_table(head))
1600                 head = ERR_PTR(-ENOENT);
1601         spin_unlock(&sysctl_lock);
1602         return head;
1603 }
1604
1605 void sysctl_head_finish(struct ctl_table_header *head)
1606 {
1607         if (!head)
1608                 return;
1609         spin_lock(&sysctl_lock);
1610         unuse_table(head);
1611         spin_unlock(&sysctl_lock);
1612 }
1613
1614 static struct ctl_table_set *
1615 lookup_header_set(struct ctl_table_root *root, struct nsproxy *namespaces)
1616 {
1617         struct ctl_table_set *set = &root->default_set;
1618         if (root->lookup)
1619                 set = root->lookup(root, namespaces);
1620         return set;
1621 }
1622
1623 static struct list_head *
1624 lookup_header_list(struct ctl_table_root *root, struct nsproxy *namespaces)
1625 {
1626         struct ctl_table_set *set = lookup_header_set(root, namespaces);
1627         return &set->list;
1628 }
1629
1630 struct ctl_table_header *__sysctl_head_next(struct nsproxy *namespaces,
1631                                             struct ctl_table_header *prev)
1632 {
1633         struct ctl_table_root *root;
1634         struct list_head *header_list;
1635         struct ctl_table_header *head;
1636         struct list_head *tmp;
1637
1638         spin_lock(&sysctl_lock);
1639         if (prev) {
1640                 head = prev;
1641                 tmp = &prev->ctl_entry;
1642                 unuse_table(prev);
1643                 goto next;
1644         }
1645         tmp = &root_table_header.ctl_entry;
1646         for (;;) {
1647                 head = list_entry(tmp, struct ctl_table_header, ctl_entry);
1648
1649                 if (!use_table(head))
1650                         goto next;
1651                 spin_unlock(&sysctl_lock);
1652                 return head;
1653         next:
1654                 root = head->root;
1655                 tmp = tmp->next;
1656                 header_list = lookup_header_list(root, namespaces);
1657                 if (tmp != header_list)
1658                         continue;
1659
1660                 do {
1661                         root = list_entry(root->root_list.next,
1662                                         struct ctl_table_root, root_list);
1663                         if (root == &sysctl_table_root)
1664                                 goto out;
1665                         header_list = lookup_header_list(root, namespaces);
1666                 } while (list_empty(header_list));
1667                 tmp = header_list->next;
1668         }
1669 out:
1670         spin_unlock(&sysctl_lock);
1671         return NULL;
1672 }
1673
1674 struct ctl_table_header *sysctl_head_next(struct ctl_table_header *prev)
1675 {
1676         return __sysctl_head_next(current->nsproxy, prev);
1677 }
1678
1679 void register_sysctl_root(struct ctl_table_root *root)
1680 {
1681         spin_lock(&sysctl_lock);
1682         list_add_tail(&root->root_list, &sysctl_table_root.root_list);
1683         spin_unlock(&sysctl_lock);
1684 }
1685
1686 /*
1687  * sysctl_perm does NOT grant the superuser all rights automatically, because
1688  * some sysctl variables are readonly even to root.
1689  */
1690
1691 static int test_perm(int mode, int op)
1692 {
1693         if (!current_euid())
1694                 mode >>= 6;
1695         else if (in_egroup_p(0))
1696                 mode >>= 3;
1697         if ((op & ~mode & (MAY_READ|MAY_WRITE|MAY_EXEC)) == 0)
1698                 return 0;
1699         return -EACCES;
1700 }
1701
1702 int sysctl_perm(struct ctl_table_root *root, struct ctl_table *table, int op)
1703 {
1704         int mode;
1705
1706         if (root->permissions)
1707                 mode = root->permissions(root, current->nsproxy, table);
1708         else
1709                 mode = table->mode;
1710
1711         return test_perm(mode, op);
1712 }
1713
1714 static void sysctl_set_parent(struct ctl_table *parent, struct ctl_table *table)
1715 {
1716         for (; table->procname; table++) {
1717                 table->parent = parent;
1718                 if (table->child)
1719                         sysctl_set_parent(table, table->child);
1720         }
1721 }
1722
1723 static __init int sysctl_init(void)
1724 {
1725         sysctl_set_parent(NULL, root_table);
1726 #ifdef CONFIG_SYSCTL_SYSCALL_CHECK
1727         sysctl_check_table(current->nsproxy, root_table);
1728 #endif
1729         return 0;
1730 }
1731
1732 core_initcall(sysctl_init);
1733
1734 static struct ctl_table *is_branch_in(struct ctl_table *branch,
1735                                       struct ctl_table *table)
1736 {
1737         struct ctl_table *p;
1738         const char *s = branch->procname;
1739
1740         /* branch should have named subdirectory as its first element */
1741         if (!s || !branch->child)
1742                 return NULL;
1743
1744         /* ... and nothing else */
1745         if (branch[1].procname)
1746                 return NULL;
1747
1748         /* table should contain subdirectory with the same name */
1749         for (p = table; p->procname; p++) {
1750                 if (!p->child)
1751                         continue;
1752                 if (p->procname && strcmp(p->procname, s) == 0)
1753                         return p;
1754         }
1755         return NULL;
1756 }
1757
1758 /* see if attaching q to p would be an improvement */
1759 static void try_attach(struct ctl_table_header *p, struct ctl_table_header *q)
1760 {
1761         struct ctl_table *to = p->ctl_table, *by = q->ctl_table;
1762         struct ctl_table *next;
1763         int is_better = 0;
1764         int not_in_parent = !p->attached_by;
1765
1766         while ((next = is_branch_in(by, to)) != NULL) {
1767                 if (by == q->attached_by)
1768                         is_better = 1;
1769                 if (to == p->attached_by)
1770                         not_in_parent = 1;
1771                 by = by->child;
1772                 to = next->child;
1773         }
1774
1775         if (is_better && not_in_parent) {
1776                 q->attached_by = by;
1777                 q->attached_to = to;
1778                 q->parent = p;
1779         }
1780 }
1781
1782 /**
1783  * __register_sysctl_paths - register a sysctl hierarchy
1784  * @root: List of sysctl headers to register on
1785  * @namespaces: Data to compute which lists of sysctl entries are visible
1786  * @path: The path to the directory the sysctl table is in.
1787  * @table: the top-level table structure
1788  *
1789  * Register a sysctl table hierarchy. @table should be a filled in ctl_table
1790  * array. A completely 0 filled entry terminates the table.
1791  *
1792  * The members of the &struct ctl_table structure are used as follows:
1793  *
1794  * procname - the name of the sysctl file under /proc/sys. Set to %NULL to not
1795  *            enter a sysctl file
1796  *
1797  * data - a pointer to data for use by proc_handler
1798  *
1799  * maxlen - the maximum size in bytes of the data
1800  *
1801  * mode - the file permissions for the /proc/sys file, and for sysctl(2)
1802  *
1803  * child - a pointer to the child sysctl table if this entry is a directory, or
1804  *         %NULL.
1805  *
1806  * proc_handler - the text handler routine (described below)
1807  *
1808  * de - for internal use by the sysctl routines
1809  *
1810  * extra1, extra2 - extra pointers usable by the proc handler routines
1811  *
1812  * Leaf nodes in the sysctl tree will be represented by a single file
1813  * under /proc; non-leaf nodes will be represented by directories.
1814  *
1815  * sysctl(2) can automatically manage read and write requests through
1816  * the sysctl table.  The data and maxlen fields of the ctl_table
1817  * struct enable minimal validation of the values being written to be
1818  * performed, and the mode field allows minimal authentication.
1819  *
1820  * There must be a proc_handler routine for any terminal nodes
1821  * mirrored under /proc/sys (non-terminals are handled by a built-in
1822  * directory handler).  Several default handlers are available to
1823  * cover common cases -
1824  *
1825  * proc_dostring(), proc_dointvec(), proc_dointvec_jiffies(),
1826  * proc_dointvec_userhz_jiffies(), proc_dointvec_minmax(),
1827  * proc_doulongvec_ms_jiffies_minmax(), proc_doulongvec_minmax()
1828  *
1829  * It is the handler's job to read the input buffer from user memory
1830  * and process it. The handler should return 0 on success.
1831  *
1832  * This routine returns %NULL on a failure to register, and a pointer
1833  * to the table header on success.
1834  */
1835 struct ctl_table_header *__register_sysctl_paths(
1836         struct ctl_table_root *root,
1837         struct nsproxy *namespaces,
1838         const struct ctl_path *path, struct ctl_table *table)
1839 {
1840         struct ctl_table_header *header;
1841         struct ctl_table *new, **prevp;
1842         unsigned int n, npath;
1843         struct ctl_table_set *set;
1844
1845         /* Count the path components */
1846         for (npath = 0; path[npath].procname; ++npath)
1847                 ;
1848
1849         /*
1850          * For each path component, allocate a 2-element ctl_table array.
1851          * The first array element will be filled with the sysctl entry
1852          * for this, the second will be the sentinel (procname == 0).
1853          *
1854          * We allocate everything in one go so that we don't have to
1855          * worry about freeing additional memory in unregister_sysctl_table.
1856          */
1857         header = kzalloc(sizeof(struct ctl_table_header) +
1858                          (2 * npath * sizeof(struct ctl_table)), GFP_KERNEL);
1859         if (!header)
1860                 return NULL;
1861
1862         new = (struct ctl_table *) (header + 1);
1863
1864         /* Now connect the dots */
1865         prevp = &header->ctl_table;
1866         for (n = 0; n < npath; ++n, ++path) {
1867                 /* Copy the procname */
1868                 new->procname = path->procname;
1869                 new->mode     = 0555;
1870
1871                 *prevp = new;
1872                 prevp = &new->child;
1873
1874                 new += 2;
1875         }
1876         *prevp = table;
1877         header->ctl_table_arg = table;
1878
1879         INIT_LIST_HEAD(&header->ctl_entry);
1880         header->used = 0;
1881         header->unregistering = NULL;
1882         header->root = root;
1883         sysctl_set_parent(NULL, header->ctl_table);
1884         header->count = 1;
1885 #ifdef CONFIG_SYSCTL_SYSCALL_CHECK
1886         if (sysctl_check_table(namespaces, header->ctl_table)) {
1887                 kfree(header);
1888                 return NULL;
1889         }
1890 #endif
1891         spin_lock(&sysctl_lock);
1892         header->set = lookup_header_set(root, namespaces);
1893         header->attached_by = header->ctl_table;
1894         header->attached_to = root_table;
1895         header->parent = &root_table_header;
1896         for (set = header->set; set; set = set->parent) {
1897                 struct ctl_table_header *p;
1898                 list_for_each_entry(p, &set->list, ctl_entry) {
1899                         if (p->unregistering)
1900                                 continue;
1901                         try_attach(p, header);
1902                 }
1903         }
1904         header->parent->count++;
1905         list_add_tail(&header->ctl_entry, &header->set->list);
1906         spin_unlock(&sysctl_lock);
1907
1908         return header;
1909 }
1910
1911 /**
1912  * register_sysctl_table_path - register a sysctl table hierarchy
1913  * @path: The path to the directory the sysctl table is in.
1914  * @table: the top-level table structure
1915  *
1916  * Register a sysctl table hierarchy. @table should be a filled in ctl_table
1917  * array. A completely 0 filled entry terminates the table.
1918  *
1919  * See __register_sysctl_paths for more details.
1920  */
1921 struct ctl_table_header *register_sysctl_paths(const struct ctl_path *path,
1922                                                 struct ctl_table *table)
1923 {
1924         return __register_sysctl_paths(&sysctl_table_root, current->nsproxy,
1925                                         path, table);
1926 }
1927
1928 /**
1929  * register_sysctl_table - register a sysctl table hierarchy
1930  * @table: the top-level table structure
1931  *
1932  * Register a sysctl table hierarchy. @table should be a filled in ctl_table
1933  * array. A completely 0 filled entry terminates the table.
1934  *
1935  * See register_sysctl_paths for more details.
1936  */
1937 struct ctl_table_header *register_sysctl_table(struct ctl_table *table)
1938 {
1939         static const struct ctl_path null_path[] = { {} };
1940
1941         return register_sysctl_paths(null_path, table);
1942 }
1943
1944 /**
1945  * unregister_sysctl_table - unregister a sysctl table hierarchy
1946  * @header: the header returned from register_sysctl_table
1947  *
1948  * Unregisters the sysctl table and all children. proc entries may not
1949  * actually be removed until they are no longer used by anyone.
1950  */
1951 void unregister_sysctl_table(struct ctl_table_header * header)
1952 {
1953         might_sleep();
1954
1955         if (header == NULL)
1956                 return;
1957
1958         spin_lock(&sysctl_lock);
1959         start_unregistering(header);
1960         if (!--header->parent->count) {
1961                 WARN_ON(1);
1962                 call_rcu(&header->parent->rcu, free_head);
1963         }
1964         if (!--header->count)
1965                 call_rcu(&header->rcu, free_head);
1966         spin_unlock(&sysctl_lock);
1967 }
1968
1969 int sysctl_is_seen(struct ctl_table_header *p)
1970 {
1971         struct ctl_table_set *set = p->set;
1972         int res;
1973         spin_lock(&sysctl_lock);
1974         if (p->unregistering)
1975                 res = 0;
1976         else if (!set->is_seen)
1977                 res = 1;
1978         else
1979                 res = set->is_seen(set);
1980         spin_unlock(&sysctl_lock);
1981         return res;
1982 }
1983
1984 void setup_sysctl_set(struct ctl_table_set *p,
1985         struct ctl_table_set *parent,
1986         int (*is_seen)(struct ctl_table_set *))
1987 {
1988         INIT_LIST_HEAD(&p->list);
1989         p->parent = parent ? parent : &sysctl_table_root.default_set;
1990         p->is_seen = is_seen;
1991 }
1992
1993 #else /* !CONFIG_SYSCTL */
1994 struct ctl_table_header *register_sysctl_table(struct ctl_table * table)
1995 {
1996         return NULL;
1997 }
1998
1999 struct ctl_table_header *register_sysctl_paths(const struct ctl_path *path,
2000                                                     struct ctl_table *table)
2001 {
2002         return NULL;
2003 }
2004
2005 void unregister_sysctl_table(struct ctl_table_header * table)
2006 {
2007 }
2008
2009 void setup_sysctl_set(struct ctl_table_set *p,
2010         struct ctl_table_set *parent,
2011         int (*is_seen)(struct ctl_table_set *))
2012 {
2013 }
2014
2015 void sysctl_head_put(struct ctl_table_header *head)
2016 {
2017 }
2018
2019 #endif /* CONFIG_SYSCTL */
2020
2021 /*
2022  * /proc/sys support
2023  */
2024
2025 #ifdef CONFIG_PROC_SYSCTL
2026
2027 static int _proc_do_string(void* data, int maxlen, int write,
2028                            void __user *buffer,
2029                            size_t *lenp, loff_t *ppos)
2030 {
2031         size_t len;
2032         char __user *p;
2033         char c;
2034
2035         if (!data || !maxlen || !*lenp) {
2036                 *lenp = 0;
2037                 return 0;
2038         }
2039
2040         if (write) {
2041                 len = 0;
2042                 p = buffer;
2043                 while (len < *lenp) {
2044                         if (get_user(c, p++))
2045                                 return -EFAULT;
2046                         if (c == 0 || c == '\n')
2047                                 break;
2048                         len++;
2049                 }
2050                 if (len >= maxlen)
2051                         len = maxlen-1;
2052                 if(copy_from_user(data, buffer, len))
2053                         return -EFAULT;
2054                 ((char *) data)[len] = 0;
2055                 *ppos += *lenp;
2056         } else {
2057                 len = strlen(data);
2058                 if (len > maxlen)
2059                         len = maxlen;
2060
2061                 if (*ppos > len) {
2062                         *lenp = 0;
2063                         return 0;
2064                 }
2065
2066                 data += *ppos;
2067                 len  -= *ppos;
2068
2069                 if (len > *lenp)
2070                         len = *lenp;
2071                 if (len)
2072                         if(copy_to_user(buffer, data, len))
2073                                 return -EFAULT;
2074                 if (len < *lenp) {
2075                         if(put_user('\n', ((char __user *) buffer) + len))
2076                                 return -EFAULT;
2077                         len++;
2078                 }
2079                 *lenp = len;
2080                 *ppos += len;
2081         }
2082         return 0;
2083 }
2084
2085 /**
2086  * proc_dostring - read a string sysctl
2087  * @table: the sysctl table
2088  * @write: %TRUE if this is a write to the sysctl file
2089  * @buffer: the user buffer
2090  * @lenp: the size of the user buffer
2091  * @ppos: file position
2092  *
2093  * Reads/writes a string from/to the user buffer. If the kernel
2094  * buffer provided is not large enough to hold the string, the
2095  * string is truncated. The copied string is %NULL-terminated.
2096  * If the string is being read by the user process, it is copied
2097  * and a newline '\n' is added. It is truncated if the buffer is
2098  * not large enough.
2099  *
2100  * Returns 0 on success.
2101  */
2102 int proc_dostring(struct ctl_table *table, int write,
2103                   void __user *buffer, size_t *lenp, loff_t *ppos)
2104 {
2105         return _proc_do_string(table->data, table->maxlen, write,
2106                                buffer, lenp, ppos);
2107 }
2108
2109 static size_t proc_skip_spaces(char **buf)
2110 {
2111         size_t ret;
2112         char *tmp = skip_spaces(*buf);
2113         ret = tmp - *buf;
2114         *buf = tmp;
2115         return ret;
2116 }
2117
2118 static void proc_skip_char(char **buf, size_t *size, const char v)
2119 {
2120         while (*size) {
2121                 if (**buf != v)
2122                         break;
2123                 (*size)--;
2124                 (*buf)++;
2125         }
2126 }
2127
2128 #define TMPBUFLEN 22
2129 /**
2130  * proc_get_long - reads an ASCII formatted integer from a user buffer
2131  *
2132  * @buf: a kernel buffer
2133  * @size: size of the kernel buffer
2134  * @val: this is where the number will be stored
2135  * @neg: set to %TRUE if number is negative
2136  * @perm_tr: a vector which contains the allowed trailers
2137  * @perm_tr_len: size of the perm_tr vector
2138  * @tr: pointer to store the trailer character
2139  *
2140  * In case of success %0 is returned and @buf and @size are updated with
2141  * the amount of bytes read. If @tr is non-NULL and a trailing
2142  * character exists (size is non-zero after returning from this
2143  * function), @tr is updated with the trailing character.
2144  */
2145 static int proc_get_long(char **buf, size_t *size,
2146                           unsigned long *val, bool *neg,
2147                           const char *perm_tr, unsigned perm_tr_len, char *tr)
2148 {
2149         int len;
2150         char *p, tmp[TMPBUFLEN];
2151
2152         if (!*size)
2153                 return -EINVAL;
2154
2155         len = *size;
2156         if (len > TMPBUFLEN - 1)
2157                 len = TMPBUFLEN - 1;
2158
2159         memcpy(tmp, *buf, len);
2160
2161         tmp[len] = 0;
2162         p = tmp;
2163         if (*p == '-' && *size > 1) {
2164                 *neg = true;
2165                 p++;
2166         } else
2167                 *neg = false;
2168         if (!isdigit(*p))
2169                 return -EINVAL;
2170
2171         *val = simple_strtoul(p, &p, 0);
2172
2173         len = p - tmp;
2174
2175         /* We don't know if the next char is whitespace thus we may accept
2176          * invalid integers (e.g. 1234...a) or two integers instead of one
2177          * (e.g. 123...1). So lets not allow such large numbers. */
2178         if (len == TMPBUFLEN - 1)
2179                 return -EINVAL;
2180
2181         if (len < *size && perm_tr_len && !memchr(perm_tr, *p, perm_tr_len))
2182                 return -EINVAL;
2183
2184         if (tr && (len < *size))
2185                 *tr = *p;
2186
2187         *buf += len;
2188         *size -= len;
2189
2190         return 0;
2191 }
2192
2193 /**
2194  * proc_put_long - converts an integer to a decimal ASCII formatted string
2195  *
2196  * @buf: the user buffer
2197  * @size: the size of the user buffer
2198  * @val: the integer to be converted
2199  * @neg: sign of the number, %TRUE for negative
2200  *
2201  * In case of success %0 is returned and @buf and @size are updated with
2202  * the amount of bytes written.
2203  */
2204 static int proc_put_long(void __user **buf, size_t *size, unsigned long val,
2205                           bool neg)
2206 {
2207         int len;
2208         char tmp[TMPBUFLEN], *p = tmp;
2209
2210         sprintf(p, "%s%lu", neg ? "-" : "", val);
2211         len = strlen(tmp);
2212         if (len > *size)
2213                 len = *size;
2214         if (copy_to_user(*buf, tmp, len))
2215                 return -EFAULT;
2216         *size -= len;
2217         *buf += len;
2218         return 0;
2219 }
2220 #undef TMPBUFLEN
2221
2222 static int proc_put_char(void __user **buf, size_t *size, char c)
2223 {
2224         if (*size) {
2225                 char __user **buffer = (char __user **)buf;
2226                 if (put_user(c, *buffer))
2227                         return -EFAULT;
2228                 (*size)--, (*buffer)++;
2229                 *buf = *buffer;
2230         }
2231         return 0;
2232 }
2233
2234 static int do_proc_dointvec_conv(bool *negp, unsigned long *lvalp,
2235                                  int *valp,
2236                                  int write, void *data)
2237 {
2238         if (write) {
2239                 *valp = *negp ? -*lvalp : *lvalp;
2240         } else {
2241                 int val = *valp;
2242                 if (val < 0) {
2243                         *negp = true;
2244                         *lvalp = (unsigned long)-val;
2245                 } else {
2246                         *negp = false;
2247                         *lvalp = (unsigned long)val;
2248                 }
2249         }
2250         return 0;
2251 }
2252
2253 static const char proc_wspace_sep[] = { ' ', '\t', '\n' };
2254
2255 static int __do_proc_dointvec(void *tbl_data, struct ctl_table *table,
2256                   int write, void __user *buffer,
2257                   size_t *lenp, loff_t *ppos,
2258                   int (*conv)(bool *negp, unsigned long *lvalp, int *valp,
2259                               int write, void *data),
2260                   void *data)
2261 {
2262         int *i, vleft, first = 1, err = 0;
2263         unsigned long page = 0;
2264         size_t left;
2265         char *kbuf;
2266
2267         if (!tbl_data || !table->maxlen || !*lenp || (*ppos && !write)) {
2268                 *lenp = 0;
2269                 return 0;
2270         }
2271
2272         i = (int *) tbl_data;
2273         vleft = table->maxlen / sizeof(*i);
2274         left = *lenp;
2275
2276         if (!conv)
2277                 conv = do_proc_dointvec_conv;
2278
2279         if (write) {
2280                 if (left > PAGE_SIZE - 1)
2281                         left = PAGE_SIZE - 1;
2282                 page = __get_free_page(GFP_TEMPORARY);
2283                 kbuf = (char *) page;
2284                 if (!kbuf)
2285                         return -ENOMEM;
2286                 if (copy_from_user(kbuf, buffer, left)) {
2287                         err = -EFAULT;
2288                         goto free;
2289                 }
2290                 kbuf[left] = 0;
2291         }
2292
2293         for (; left && vleft--; i++, first=0) {
2294                 unsigned long lval;
2295                 bool neg;
2296
2297                 if (write) {
2298                         left -= proc_skip_spaces(&kbuf);
2299
2300                         if (!left)
2301                                 break;
2302                         err = proc_get_long(&kbuf, &left, &lval, &neg,
2303                                              proc_wspace_sep,
2304                                              sizeof(proc_wspace_sep), NULL);
2305                         if (err)
2306                                 break;
2307                         if (conv(&neg, &lval, i, 1, data)) {
2308                                 err = -EINVAL;
2309                                 break;
2310                         }
2311                 } else {
2312                         if (conv(&neg, &lval, i, 0, data)) {
2313                                 err = -EINVAL;
2314                                 break;
2315                         }
2316                         if (!first)
2317                                 err = proc_put_char(&buffer, &left, '\t');
2318                         if (err)
2319                                 break;
2320                         err = proc_put_long(&buffer, &left, lval, neg);
2321                         if (err)
2322                                 break;
2323                 }
2324         }
2325
2326         if (!write && !first && left && !err)
2327                 err = proc_put_char(&buffer, &left, '\n');
2328         if (write && !err && left)
2329                 left -= proc_skip_spaces(&kbuf);
2330 free:
2331         if (write) {
2332                 free_page(page);
2333                 if (first)
2334                         return err ? : -EINVAL;
2335         }
2336         *lenp -= left;
2337         *ppos += *lenp;
2338         return err;
2339 }
2340
2341 static int do_proc_dointvec(struct ctl_table *table, int write,
2342                   void __user *buffer, size_t *lenp, loff_t *ppos,
2343                   int (*conv)(bool *negp, unsigned long *lvalp, int *valp,
2344                               int write, void *data),
2345                   void *data)
2346 {
2347         return __do_proc_dointvec(table->data, table, write,
2348                         buffer, lenp, ppos, conv, data);
2349 }
2350
2351 /**
2352  * proc_dointvec - read a vector of integers
2353  * @table: the sysctl table
2354  * @write: %TRUE if this is a write to the sysctl file
2355  * @buffer: the user buffer
2356  * @lenp: the size of the user buffer
2357  * @ppos: file position
2358  *
2359  * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2360  * values from/to the user buffer, treated as an ASCII string.
2361  *
2362  * Returns 0 on success.
2363  */
2364 int proc_dointvec(struct ctl_table *table, int write,
2365                      void __user *buffer, size_t *lenp, loff_t *ppos)
2366 {
2367     return do_proc_dointvec(table,write,buffer,lenp,ppos,
2368                             NULL,NULL);
2369 }
2370
2371 /*
2372  * Taint values can only be increased
2373  * This means we can safely use a temporary.
2374  */
2375 static int proc_taint(struct ctl_table *table, int write,
2376                                void __user *buffer, size_t *lenp, loff_t *ppos)
2377 {
2378         struct ctl_table t;
2379         unsigned long tmptaint = get_taint();
2380         int err;
2381
2382         if (write && !capable(CAP_SYS_ADMIN))
2383                 return -EPERM;
2384
2385         t = *table;
2386         t.data = &tmptaint;
2387         err = proc_doulongvec_minmax(&t, write, buffer, lenp, ppos);
2388         if (err < 0)
2389                 return err;
2390
2391         if (write) {
2392                 /*
2393                  * Poor man's atomic or. Not worth adding a primitive
2394                  * to everyone's atomic.h for this
2395                  */
2396                 int i;
2397                 for (i = 0; i < BITS_PER_LONG && tmptaint >> i; i++) {
2398                         if ((tmptaint >> i) & 1)
2399                                 add_taint(i);
2400                 }
2401         }
2402
2403         return err;
2404 }
2405
2406 #ifdef CONFIG_PRINTK
2407 static int proc_dmesg_restrict(struct ctl_table *table, int write,
2408                                 void __user *buffer, size_t *lenp, loff_t *ppos)
2409 {
2410         if (write && !capable(CAP_SYS_ADMIN))
2411                 return -EPERM;
2412
2413         return proc_dointvec_minmax(table, write, buffer, lenp, ppos);
2414 }
2415 #endif
2416
2417 struct do_proc_dointvec_minmax_conv_param {
2418         int *min;
2419         int *max;
2420 };
2421
2422 static int do_proc_dointvec_minmax_conv(bool *negp, unsigned long *lvalp,
2423                                         int *valp,
2424                                         int write, void *data)
2425 {
2426         struct do_proc_dointvec_minmax_conv_param *param = data;
2427         if (write) {
2428                 int val = *negp ? -*lvalp : *lvalp;
2429                 if ((param->min && *param->min > val) ||
2430                     (param->max && *param->max < val))
2431                         return -EINVAL;
2432                 *valp = val;
2433         } else {
2434                 int val = *valp;
2435                 if (val < 0) {
2436                         *negp = true;
2437                         *lvalp = (unsigned long)-val;
2438                 } else {
2439                         *negp = false;
2440                         *lvalp = (unsigned long)val;
2441                 }
2442         }
2443         return 0;
2444 }
2445
2446 /**
2447  * proc_dointvec_minmax - read a vector of integers with min/max values
2448  * @table: the sysctl table
2449  * @write: %TRUE if this is a write to the sysctl file
2450  * @buffer: the user buffer
2451  * @lenp: the size of the user buffer
2452  * @ppos: file position
2453  *
2454  * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2455  * values from/to the user buffer, treated as an ASCII string.
2456  *
2457  * This routine will ensure the values are within the range specified by
2458  * table->extra1 (min) and table->extra2 (max).
2459  *
2460  * Returns 0 on success.
2461  */
2462 int proc_dointvec_minmax(struct ctl_table *table, int write,
2463                   void __user *buffer, size_t *lenp, loff_t *ppos)
2464 {
2465         struct do_proc_dointvec_minmax_conv_param param = {
2466                 .min = (int *) table->extra1,
2467                 .max = (int *) table->extra2,
2468         };
2469         return do_proc_dointvec(table, write, buffer, lenp, ppos,
2470                                 do_proc_dointvec_minmax_conv, &param);
2471 }
2472
2473 static int __do_proc_doulongvec_minmax(void *data, struct ctl_table *table, int write,
2474                                      void __user *buffer,
2475                                      size_t *lenp, loff_t *ppos,
2476                                      unsigned long convmul,
2477                                      unsigned long convdiv)
2478 {
2479         unsigned long *i, *min, *max;
2480         int vleft, first = 1, err = 0;
2481         unsigned long page = 0;
2482         size_t left;
2483         char *kbuf;
2484
2485         if (!data || !table->maxlen || !*lenp || (*ppos && !write)) {
2486                 *lenp = 0;
2487                 return 0;
2488         }
2489
2490         i = (unsigned long *) data;
2491         min = (unsigned long *) table->extra1;
2492         max = (unsigned long *) table->extra2;
2493         vleft = table->maxlen / sizeof(unsigned long);
2494         left = *lenp;
2495
2496         if (write) {
2497                 if (left > PAGE_SIZE - 1)
2498                         left = PAGE_SIZE - 1;
2499                 page = __get_free_page(GFP_TEMPORARY);
2500                 kbuf = (char *) page;
2501                 if (!kbuf)
2502                         return -ENOMEM;
2503                 if (copy_from_user(kbuf, buffer, left)) {
2504                         err = -EFAULT;
2505                         goto free;
2506                 }
2507                 kbuf[left] = 0;
2508         }
2509
2510         for (; left && vleft--; i++, first = 0) {
2511                 unsigned long val;
2512
2513                 if (write) {
2514                         bool neg;
2515
2516                         left -= proc_skip_spaces(&kbuf);
2517
2518                         err = proc_get_long(&kbuf, &left, &val, &neg,
2519                                              proc_wspace_sep,
2520                                              sizeof(proc_wspace_sep), NULL);
2521                         if (err)
2522                                 break;
2523                         if (neg)
2524                                 continue;
2525                         if ((min && val < *min) || (max && val > *max))
2526                                 continue;
2527                         *i = val;
2528                 } else {
2529                         val = convdiv * (*i) / convmul;
2530                         if (!first)
2531                                 err = proc_put_char(&buffer, &left, '\t');
2532                         err = proc_put_long(&buffer, &left, val, false);
2533                         if (err)
2534                                 break;
2535                 }
2536         }
2537
2538         if (!write && !first && left && !err)
2539                 err = proc_put_char(&buffer, &left, '\n');
2540         if (write && !err)
2541                 left -= proc_skip_spaces(&kbuf);
2542 free:
2543         if (write) {
2544                 free_page(page);
2545                 if (first)
2546                         return err ? : -EINVAL;
2547         }
2548         *lenp -= left;
2549         *ppos += *lenp;
2550         return err;
2551 }
2552
2553 static int do_proc_doulongvec_minmax(struct ctl_table *table, int write,
2554                                      void __user *buffer,
2555                                      size_t *lenp, loff_t *ppos,
2556                                      unsigned long convmul,
2557                                      unsigned long convdiv)
2558 {
2559         return __do_proc_doulongvec_minmax(table->data, table, write,
2560                         buffer, lenp, ppos, convmul, convdiv);
2561 }
2562
2563 /**
2564  * proc_doulongvec_minmax - read a vector of long integers with min/max values
2565  * @table: the sysctl table
2566  * @write: %TRUE if this is a write to the sysctl file
2567  * @buffer: the user buffer
2568  * @lenp: the size of the user buffer
2569  * @ppos: file position
2570  *
2571  * Reads/writes up to table->maxlen/sizeof(unsigned long) unsigned long
2572  * values from/to the user buffer, treated as an ASCII string.
2573  *
2574  * This routine will ensure the values are within the range specified by
2575  * table->extra1 (min) and table->extra2 (max).
2576  *
2577  * Returns 0 on success.
2578  */
2579 int proc_doulongvec_minmax(struct ctl_table *table, int write,
2580                            void __user *buffer, size_t *lenp, loff_t *ppos)
2581 {
2582     return do_proc_doulongvec_minmax(table, write, buffer, lenp, ppos, 1l, 1l);
2583 }
2584
2585 /**
2586  * proc_doulongvec_ms_jiffies_minmax - read a vector of millisecond values with min/max values
2587  * @table: the sysctl table
2588  * @write: %TRUE if this is a write to the sysctl file
2589  * @buffer: the user buffer
2590  * @lenp: the size of the user buffer
2591  * @ppos: file position
2592  *
2593  * Reads/writes up to table->maxlen/sizeof(unsigned long) unsigned long
2594  * values from/to the user buffer, treated as an ASCII string. The values
2595  * are treated as milliseconds, and converted to jiffies when they are stored.
2596  *
2597  * This routine will ensure the values are within the range specified by
2598  * table->extra1 (min) and table->extra2 (max).
2599  *
2600  * Returns 0 on success.
2601  */
2602 int proc_doulongvec_ms_jiffies_minmax(struct ctl_table *table, int write,
2603                                       void __user *buffer,
2604                                       size_t *lenp, loff_t *ppos)
2605 {
2606     return do_proc_doulongvec_minmax(table, write, buffer,
2607                                      lenp, ppos, HZ, 1000l);
2608 }
2609
2610
2611 static int do_proc_dointvec_jiffies_conv(bool *negp, unsigned long *lvalp,
2612                                          int *valp,
2613                                          int write, void *data)
2614 {
2615         if (write) {
2616                 if (*lvalp > LONG_MAX / HZ)
2617                         return 1;
2618                 *valp = *negp ? -(*lvalp*HZ) : (*lvalp*HZ);
2619         } else {
2620                 int val = *valp;
2621                 unsigned long lval;
2622                 if (val < 0) {
2623                         *negp = true;
2624                         lval = (unsigned long)-val;
2625                 } else {
2626                         *negp = false;
2627                         lval = (unsigned long)val;
2628                 }
2629                 *lvalp = lval / HZ;
2630         }
2631         return 0;
2632 }
2633
2634 static int do_proc_dointvec_userhz_jiffies_conv(bool *negp, unsigned long *lvalp,
2635                                                 int *valp,
2636                                                 int write, void *data)
2637 {
2638         if (write) {
2639                 if (USER_HZ < HZ && *lvalp > (LONG_MAX / HZ) * USER_HZ)
2640                         return 1;
2641                 *valp = clock_t_to_jiffies(*negp ? -*lvalp : *lvalp);
2642         } else {
2643                 int val = *valp;
2644                 unsigned long lval;
2645                 if (val < 0) {
2646                         *negp = true;
2647                         lval = (unsigned long)-val;
2648                 } else {
2649                         *negp = false;
2650                         lval = (unsigned long)val;
2651                 }
2652                 *lvalp = jiffies_to_clock_t(lval);
2653         }
2654         return 0;
2655 }
2656
2657 static int do_proc_dointvec_ms_jiffies_conv(bool *negp, unsigned long *lvalp,
2658                                             int *valp,
2659                                             int write, void *data)
2660 {
2661         if (write) {
2662                 *valp = msecs_to_jiffies(*negp ? -*lvalp : *lvalp);
2663         } else {
2664                 int val = *valp;
2665                 unsigned long lval;
2666                 if (val < 0) {
2667                         *negp = true;
2668                         lval = (unsigned long)-val;
2669                 } else {
2670                         *negp = false;
2671                         lval = (unsigned long)val;
2672                 }
2673                 *lvalp = jiffies_to_msecs(lval);
2674         }
2675         return 0;
2676 }
2677
2678 /**
2679  * proc_dointvec_jiffies - read a vector of integers as seconds
2680  * @table: the sysctl table
2681  * @write: %TRUE if this is a write to the sysctl file
2682  * @buffer: the user buffer
2683  * @lenp: the size of the user buffer
2684  * @ppos: file position
2685  *
2686  * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2687  * values from/to the user buffer, treated as an ASCII string.
2688  * The values read are assumed to be in seconds, and are converted into
2689  * jiffies.
2690  *
2691  * Returns 0 on success.
2692  */
2693 int proc_dointvec_jiffies(struct ctl_table *table, int write,
2694                           void __user *buffer, size_t *lenp, loff_t *ppos)
2695 {
2696     return do_proc_dointvec(table,write,buffer,lenp,ppos,
2697                             do_proc_dointvec_jiffies_conv,NULL);
2698 }
2699
2700 /**
2701  * proc_dointvec_userhz_jiffies - read a vector of integers as 1/USER_HZ seconds
2702  * @table: the sysctl table
2703  * @write: %TRUE if this is a write to the sysctl file
2704  * @buffer: the user buffer
2705  * @lenp: the size of the user buffer
2706  * @ppos: pointer to the file position
2707  *
2708  * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2709  * values from/to the user buffer, treated as an ASCII string.
2710  * The values read are assumed to be in 1/USER_HZ seconds, and
2711  * are converted into jiffies.
2712  *
2713  * Returns 0 on success.
2714  */
2715 int proc_dointvec_userhz_jiffies(struct ctl_table *table, int write,
2716                                  void __user *buffer, size_t *lenp, loff_t *ppos)
2717 {
2718     return do_proc_dointvec(table,write,buffer,lenp,ppos,
2719                             do_proc_dointvec_userhz_jiffies_conv,NULL);
2720 }
2721
2722 /**
2723  * proc_dointvec_ms_jiffies - read a vector of integers as 1 milliseconds
2724  * @table: the sysctl table
2725  * @write: %TRUE if this is a write to the sysctl file
2726  * @buffer: the user buffer
2727  * @lenp: the size of the user buffer
2728  * @ppos: file position
2729  * @ppos: the current position in the file
2730  *
2731  * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2732  * values from/to the user buffer, treated as an ASCII string.
2733  * The values read are assumed to be in 1/1000 seconds, and
2734  * are converted into jiffies.
2735  *
2736  * Returns 0 on success.
2737  */
2738 int proc_dointvec_ms_jiffies(struct ctl_table *table, int write,
2739                              void __user *buffer, size_t *lenp, loff_t *ppos)
2740 {
2741         return do_proc_dointvec(table, write, buffer, lenp, ppos,
2742                                 do_proc_dointvec_ms_jiffies_conv, NULL);
2743 }
2744
2745 static int proc_do_cad_pid(struct ctl_table *table, int write,
2746                            void __user *buffer, size_t *lenp, loff_t *ppos)
2747 {
2748         struct pid *new_pid;
2749         pid_t tmp;
2750         int r;
2751
2752         tmp = pid_vnr(cad_pid);
2753
2754         r = __do_proc_dointvec(&tmp, table, write, buffer,
2755                                lenp, ppos, NULL, NULL);
2756         if (r || !write)
2757                 return r;
2758
2759         new_pid = find_get_pid(tmp);
2760         if (!new_pid)
2761                 return -ESRCH;
2762
2763         put_pid(xchg(&cad_pid, new_pid));
2764         return 0;
2765 }
2766
2767 /**
2768  * proc_do_large_bitmap - read/write from/to a large bitmap
2769  * @table: the sysctl table
2770  * @write: %TRUE if this is a write to the sysctl file
2771  * @buffer: the user buffer
2772  * @lenp: the size of the user buffer
2773  * @ppos: file position
2774  *
2775  * The bitmap is stored at table->data and the bitmap length (in bits)
2776  * in table->maxlen.
2777  *
2778  * We use a range comma separated format (e.g. 1,3-4,10-10) so that
2779  * large bitmaps may be represented in a compact manner. Writing into
2780  * the file will clear the bitmap then update it with the given input.
2781  *
2782  * Returns 0 on success.
2783  */
2784 int proc_do_large_bitmap(struct ctl_table *table, int write,
2785                          void __user *buffer, size_t *lenp, loff_t *ppos)
2786 {
2787         int err = 0;
2788         bool first = 1;
2789         size_t left = *lenp;
2790         unsigned long bitmap_len = table->maxlen;
2791         unsigned long *bitmap = (unsigned long *) table->data;
2792         unsigned long *tmp_bitmap = NULL;
2793         char tr_a[] = { '-', ',', '\n' }, tr_b[] = { ',', '\n', 0 }, c;
2794
2795         if (!bitmap_len || !left || (*ppos && !write)) {
2796                 *lenp = 0;
2797                 return 0;
2798         }
2799
2800         if (write) {
2801                 unsigned long page = 0;
2802                 char *kbuf;
2803
2804                 if (left > PAGE_SIZE - 1)
2805                         left = PAGE_SIZE - 1;
2806
2807                 page = __get_free_page(GFP_TEMPORARY);
2808                 kbuf = (char *) page;
2809                 if (!kbuf)
2810                         return -ENOMEM;
2811                 if (copy_from_user(kbuf, buffer, left)) {
2812                         free_page(page);
2813                         return -EFAULT;
2814                 }
2815                 kbuf[left] = 0;
2816
2817                 tmp_bitmap = kzalloc(BITS_TO_LONGS(bitmap_len) * sizeof(unsigned long),
2818                                      GFP_KERNEL);
2819                 if (!tmp_bitmap) {
2820                         free_page(page);
2821                         return -ENOMEM;
2822                 }
2823                 proc_skip_char(&kbuf, &left, '\n');
2824                 while (!err && left) {
2825                         unsigned long val_a, val_b;
2826                         bool neg;
2827
2828                         err = proc_get_long(&kbuf, &left, &val_a, &neg, tr_a,
2829                                              sizeof(tr_a), &c);
2830                         if (err)
2831                                 break;
2832                         if (val_a >= bitmap_len || neg) {
2833                                 err = -EINVAL;
2834                                 break;
2835                         }
2836
2837                         val_b = val_a;
2838                         if (left) {
2839                                 kbuf++;
2840                                 left--;
2841                         }
2842
2843                         if (c == '-') {
2844                                 err = proc_get_long(&kbuf, &left, &val_b,
2845                                                      &neg, tr_b, sizeof(tr_b),
2846                                                      &c);
2847                                 if (err)
2848                                         break;
2849                                 if (val_b >= bitmap_len || neg ||
2850                                     val_a > val_b) {
2851                                         err = -EINVAL;
2852                                         break;
2853                                 }
2854                                 if (left) {
2855                                         kbuf++;
2856                                         left--;
2857                                 }
2858                         }
2859
2860                         while (val_a <= val_b)
2861                                 set_bit(val_a++, tmp_bitmap);
2862
2863                         first = 0;
2864                         proc_skip_char(&kbuf, &left, '\n');
2865                 }
2866                 free_page(page);
2867         } else {
2868                 unsigned long bit_a, bit_b = 0;
2869
2870                 while (left) {
2871                         bit_a = find_next_bit(bitmap, bitmap_len, bit_b);
2872                         if (bit_a >= bitmap_len)
2873                                 break;
2874                         bit_b = find_next_zero_bit(bitmap, bitmap_len,
2875                                                    bit_a + 1) - 1;
2876
2877                         if (!first) {
2878                                 err = proc_put_char(&buffer, &left, ',');
2879                                 if (err)
2880                                         break;
2881                         }
2882                         err = proc_put_long(&buffer, &left, bit_a, false);
2883                         if (err)
2884                                 break;
2885                         if (bit_a != bit_b) {
2886                                 err = proc_put_char(&buffer, &left, '-');
2887                                 if (err)
2888                                         break;
2889                                 err = proc_put_long(&buffer, &left, bit_b, false);
2890                                 if (err)
2891                                         break;
2892                         }
2893
2894                         first = 0; bit_b++;
2895                 }
2896                 if (!err)
2897                         err = proc_put_char(&buffer, &left, '\n');
2898         }
2899
2900         if (!err) {
2901                 if (write) {
2902                         if (*ppos)
2903                                 bitmap_or(bitmap, bitmap, tmp_bitmap, bitmap_len);
2904                         else
2905                                 memcpy(bitmap, tmp_bitmap,
2906                                         BITS_TO_LONGS(bitmap_len) * sizeof(unsigned long));
2907                 }
2908                 kfree(tmp_bitmap);
2909                 *lenp -= left;
2910                 *ppos += *lenp;
2911                 return 0;
2912         } else {
2913                 kfree(tmp_bitmap);
2914                 return err;
2915         }
2916 }
2917
2918 #else /* CONFIG_PROC_SYSCTL */
2919
2920 int proc_dostring(struct ctl_table *table, int write,
2921                   void __user *buffer, size_t *lenp, loff_t *ppos)
2922 {
2923         return -ENOSYS;
2924 }
2925
2926 int proc_dointvec(struct ctl_table *table, int write,
2927                   void __user *buffer, size_t *lenp, loff_t *ppos)
2928 {
2929         return -ENOSYS;
2930 }
2931
2932 int proc_dointvec_minmax(struct ctl_table *table, int write,
2933                     void __user *buffer, size_t *lenp, loff_t *ppos)
2934 {
2935         return -ENOSYS;
2936 }
2937
2938 int proc_dointvec_jiffies(struct ctl_table *table, int write,
2939                     void __user *buffer, size_t *lenp, loff_t *ppos)
2940 {
2941         return -ENOSYS;
2942 }
2943
2944 int proc_dointvec_userhz_jiffies(struct ctl_table *table, int write,
2945                     void __user *buffer, size_t *lenp, loff_t *ppos)
2946 {
2947         return -ENOSYS;
2948 }
2949
2950 int proc_dointvec_ms_jiffies(struct ctl_table *table, int write,
2951                              void __user *buffer, size_t *lenp, loff_t *ppos)
2952 {
2953         return -ENOSYS;
2954 }
2955
2956 int proc_doulongvec_minmax(struct ctl_table *table, int write,
2957                     void __user *buffer, size_t *lenp, loff_t *ppos)
2958 {
2959         return -ENOSYS;
2960 }
2961
2962 int proc_doulongvec_ms_jiffies_minmax(struct ctl_table *table, int write,
2963                                       void __user *buffer,
2964                                       size_t *lenp, loff_t *ppos)
2965 {
2966     return -ENOSYS;
2967 }
2968
2969
2970 #endif /* CONFIG_PROC_SYSCTL */
2971
2972 /*
2973  * No sense putting this after each symbol definition, twice,
2974  * exception granted :-)
2975  */
2976 EXPORT_SYMBOL(proc_dointvec);
2977 EXPORT_SYMBOL(proc_dointvec_jiffies);
2978 EXPORT_SYMBOL(proc_dointvec_minmax);
2979 EXPORT_SYMBOL(proc_dointvec_userhz_jiffies);
2980 EXPORT_SYMBOL(proc_dointvec_ms_jiffies);
2981 EXPORT_SYMBOL(proc_dostring);
2982 EXPORT_SYMBOL(proc_doulongvec_minmax);
2983 EXPORT_SYMBOL(proc_doulongvec_ms_jiffies_minmax);
2984 EXPORT_SYMBOL(register_sysctl_table);
2985 EXPORT_SYMBOL(register_sysctl_paths);
2986 EXPORT_SYMBOL(unregister_sysctl_table);