kernel/sysctl.c

   1 /*
   2  * sysctl.c: General linux system control interface
   3  *
   4  * Begun 24 March 1995, Stephen Tweedie
   5  * Added /proc support, Dec 1995
   6  * Added bdflush entry and intvec min/max checking, 2/23/96, Tom Dyas.
   7  * Added hooks for /proc/sys/net (minor, minor patch), 96/4/1, Mike Shaver.
   8  * Added kernel/java-{interpreter,appletviewer}, 96/5/10, Mike Shaver.
   9  * Dynamic registration fixes, Stephen Tweedie.
  10  * Added kswapd-interval, ctrl-alt-del, printk stuff, 1/8/97, Chris Horn.
  11  * Made sysctl support optional via CONFIG_SYSCTL, 1/10/97, Chris
  12  *  Horn.
  13  * Added proc_doulongvec_ms_jiffies_minmax, 09/08/99, Carlos H. Bauer.
  14  * Added proc_doulongvec_minmax, 09/08/99, Carlos H. Bauer.
  15  * Changed linked lists to use list.h instead of lists.h, 02/24/00, Bill
  16  *  Wendling.
  17  * The list_for_each() macro wasn't appropriate for the sysctl loop.
  18  *  Removed it and replaced it with older style, 03/23/00, Bill Wendling
  19  */
  20
  21 #include <linux/module.h>
  22 #include <linux/mm.h>
  23 #include <linux/swap.h>
  24 #include <linux/slab.h>
  25 #include <linux/sysctl.h>
  26 #include <linux/signal.h>
  27 #include <linux/printk.h>
  28 #include <linux/proc_fs.h>
  29 #include <linux/security.h>
  30 #include <linux/ctype.h>
  31 #include <linux/kmemcheck.h>
  32 #include <linux/fs.h>
  33 #include <linux/init.h>
  34 #include <linux/kernel.h>
  35 #include <linux/kobject.h>
  36 #include <linux/net.h>
  37 #include <linux/sysrq.h>
  38 #include <linux/highuid.h>
  39 #include <linux/writeback.h>
  40 #include <linux/ratelimit.h>
  41 #include <linux/compaction.h>
  42 #include <linux/hugetlb.h>
  43 #include <linux/initrd.h>
  44 #include <linux/key.h>
  45 #include <linux/times.h>
  46 #include <linux/limits.h>
  47 #include <linux/dcache.h>
  48 #include <linux/dnotify.h>
  49 #include <linux/syscalls.h>
  50 #include <linux/vmstat.h>
  51 #include <linux/nfs_fs.h>
  52 #include <linux/acpi.h>
  53 #include <linux/reboot.h>
  54 #include <linux/ftrace.h>
  55 #include <linux/perf_event.h>
  56 #include <linux/kprobes.h>
  57 #include <linux/pipe_fs_i.h>
  58 #include <linux/oom.h>
  59
  60 #include <asm/uaccess.h>
  61 #include <asm/processor.h>
  62
  63 #ifdef CONFIG_X86
  64 #include <asm/nmi.h>
  65 #include <asm/stacktrace.h>
  66 #include <asm/io.h>
  67 #endif
  68 #ifdef CONFIG_BSD_PROCESS_ACCT
  69 #include <linux/acct.h>
  70 #endif
  71 #ifdef CONFIG_RT_MUTEXES
  72 #include <linux/rtmutex.h>
  73 #endif
  74 #if defined(CONFIG_PROVE_LOCKING) || defined(CONFIG_LOCK_STAT)
  75 #include <linux/lockdep.h>
  76 #endif
  77 #ifdef CONFIG_CHR_DEV_SG
  78 #include <scsi/sg.h>
  79 #endif
  80
  81 #ifdef CONFIG_LOCKUP_DETECTOR
  82 #include <linux/nmi.h>
  83 #endif
  84
  85
  86 #if defined(CONFIG_SYSCTL)
  87
  88 /* External variables not in a header file. */
  89 extern int sysctl_overcommit_memory;
  90 extern int sysctl_overcommit_ratio;
  91 extern int max_threads;
  92 extern int core_uses_pid;
  93 extern int suid_dumpable;
  94 extern char core_pattern[];
  95 extern unsigned int core_pipe_limit;
  96 extern int pid_max;
  97 extern int min_free_kbytes;
  98 extern int pid_max_min, pid_max_max;
  99 extern int sysctl_drop_caches;
 100 extern int percpu_pagelist_fraction;
 101 extern int compat_log;
 102 extern int latencytop_enabled;
 103 extern int sysctl_nr_open_min, sysctl_nr_open_max;
 104 #ifndef CONFIG_MMU
 105 extern int sysctl_nr_trim_pages;
 106 #endif
 107 #ifdef CONFIG_BLOCK
 108 extern int blk_iopoll_enabled;
 109 #endif
 110
 111 /* Constants used for minimum and  maximum */
 112 #ifdef CONFIG_LOCKUP_DETECTOR
 113 static int sixty = 60;
 114 static int neg_one = -1;
 115 #endif
 116
 117 static int zero;
 118 static int __maybe_unused one = 1;
 119 static int __maybe_unused two = 2;
 120 static unsigned long one_ul = 1;
 121 static int one_hundred = 100;
 122 #ifdef CONFIG_PRINTK
 123 static int ten_thousand = 10000;
 124 #endif
 125
 126 /* this is needed for the proc_doulongvec_minmax of vm_dirty_bytes */
 127 static unsigned long dirty_bytes_min = 2 * PAGE_SIZE;
 128
 129 /* this is needed for the proc_dointvec_minmax for [fs_]overflow UID and GID */
 130 static int maxolduid = 65535;
 131 static int minolduid;
 132 static int min_percpu_pagelist_fract = 8;
 133
 134 static int ngroups_max = NGROUPS_MAX;
 135
 136 #ifdef CONFIG_INOTIFY_USER
 137 #include <linux/inotify.h>
 138 #endif
 139 #ifdef CONFIG_SPARC
 140 #include <asm/system.h>
 141 #endif
 142
 143 #ifdef CONFIG_SPARC64
 144 extern int sysctl_tsb_ratio;
 145 #endif
 146
 147 #ifdef __hppa__
 148 extern int pwrsw_enabled;
 149 extern int unaligned_enabled;
 150 #endif
 151
 152 #ifdef CONFIG_S390
 153 #ifdef CONFIG_MATHEMU
 154 extern int sysctl_ieee_emulation_warnings;
 155 #endif
 156 extern int sysctl_userprocess_debug;
 157 extern int spin_retry;
 158 #endif
 159
 160 #ifdef CONFIG_IA64
 161 extern int no_unaligned_warning;
 162 extern int unaligned_dump_stack;
 163 #endif
 164
 165 #ifdef CONFIG_PROC_SYSCTL
 166 static int proc_do_cad_pid(struct ctl_table *table, int write,
 167                   void __user *buffer, size_t *lenp, loff_t *ppos);
 168 static int proc_taint(struct ctl_table *table, int write,
 169                                void __user *buffer, size_t *lenp, loff_t *ppos);
 170 #endif
 171
 172 #ifdef CONFIG_MAGIC_SYSRQ
 173 /* Note: sysrq code uses it's own private copy */
 174 static int __sysrq_enabled = SYSRQ_DEFAULT_ENABLE;
 175
 176 static int sysrq_sysctl_handler(ctl_table *table, int write,
 177                                 void __user *buffer, size_t *lenp,
 178                                 loff_t *ppos)
 179 {
 180         int error;
 181
 182         error = proc_dointvec(table, write, buffer, lenp, ppos);
 183         if (error)
 184                 return error;
 185
 186         if (write)
 187                 sysrq_toggle_support(__sysrq_enabled);
 188
 189         return 0;
 190 }
 191
 192 #endif
 193
 194 static struct ctl_table root_table[];
 195 static struct ctl_table_root sysctl_table_root;
 196 static struct ctl_table_header root_table_header = {
 197         .count = 1,
 198         .ctl_table = root_table,
 199         .ctl_entry = LIST_HEAD_INIT(sysctl_table_root.default_set.list),
 200         .root = &sysctl_table_root,
 201         .set = &sysctl_table_root.default_set,
 202 };
 203 static struct ctl_table_root sysctl_table_root = {
 204         .root_list = LIST_HEAD_INIT(sysctl_table_root.root_list),
 205         .default_set.list = LIST_HEAD_INIT(root_table_header.ctl_entry),
 206 };
 207
 208 static struct ctl_table kern_table[];
 209 static struct ctl_table vm_table[];
 210 static struct ctl_table fs_table[];
 211 static struct ctl_table debug_table[];
 212 static struct ctl_table dev_table[];
 213 extern struct ctl_table random_table[];
 214 #ifdef CONFIG_EPOLL
 215 extern struct ctl_table epoll_table[];
 216 #endif
 217
 218 #ifdef HAVE_ARCH_PICK_MMAP_LAYOUT
 219 int sysctl_legacy_va_layout;
 220 #endif
 221
 222 /* The default sysctl tables: */
 223
 224 static struct ctl_table root_table[] = {
 225         {
 226                 .procname       = "kernel",
 227                 .mode           = 0555,
 228                 .child          = kern_table,
 229         },
 230         {
 231                 .procname       = "vm",
 232                 .mode           = 0555,
 233                 .child          = vm_table,
 234         },
 235         {
 236                 .procname       = "fs",
 237                 .mode           = 0555,
 238                 .child          = fs_table,
 239         },
 240         {
 241                 .procname       = "debug",
 242                 .mode           = 0555,
 243                 .child          = debug_table,
 244         },
 245         {
 246                 .procname       = "dev",
 247                 .mode           = 0555,
 248                 .child          = dev_table,
 249         },
 250         { }
 251 };
 252
 253 #ifdef CONFIG_SCHED_DEBUG
 254 static int min_sched_granularity_ns = 100000;           /* 100 usecs */
 255 static int max_sched_granularity_ns = NSEC_PER_SEC;     /* 1 second */
 256 static int min_wakeup_granularity_ns;                   /* 0 usecs */
 257 static int max_wakeup_granularity_ns = NSEC_PER_SEC;    /* 1 second */
 258 static int min_sched_tunable_scaling = SCHED_TUNABLESCALING_NONE;
 259 static int max_sched_tunable_scaling = SCHED_TUNABLESCALING_END-1;
 260 #endif
 261
 262 #ifdef CONFIG_COMPACTION
 263 static int min_extfrag_threshold;
 264 static int max_extfrag_threshold = 1000;
 265 #endif
 266
 267 static struct ctl_table kern_table[] = {
 268         {
 269                 .procname       = "sched_child_runs_first",
 270                 .data           = &sysctl_sched_child_runs_first,
 271                 .maxlen         = sizeof(unsigned int),
 272                 .mode           = 0644,
 273                 .proc_handler   = proc_dointvec,
 274         },
 275 #ifdef CONFIG_SCHED_DEBUG
 276         {
 277                 .procname       = "sched_min_granularity_ns",
 278                 .data           = &sysctl_sched_min_granularity,
 279                 .maxlen         = sizeof(unsigned int),
 280                 .mode           = 0644,
 281                 .proc_handler   = sched_proc_update_handler,
 282                 .extra1         = &min_sched_granularity_ns,
 283                 .extra2         = &max_sched_granularity_ns,
 284         },
 285         {
 286                 .procname       = "sched_latency_ns",
 287                 .data           = &sysctl_sched_latency,
 288                 .maxlen         = sizeof(unsigned int),
 289                 .mode           = 0644,
 290                 .proc_handler   = sched_proc_update_handler,
 291                 .extra1         = &min_sched_granularity_ns,
 292                 .extra2         = &max_sched_granularity_ns,
 293         },
 294         {
 295                 .procname       = "sched_wakeup_granularity_ns",
 296                 .data           = &sysctl_sched_wakeup_granularity,
 297                 .maxlen         = sizeof(unsigned int),
 298                 .mode           = 0644,
 299                 .proc_handler   = sched_proc_update_handler,
 300                 .extra1         = &min_wakeup_granularity_ns,
 301                 .extra2         = &max_wakeup_granularity_ns,
 302         },
 303         {
 304                 .procname       = "sched_tunable_scaling",
 305                 .data           = &sysctl_sched_tunable_scaling,
 306                 .maxlen         = sizeof(enum sched_tunable_scaling),
 307                 .mode           = 0644,
 308                 .proc_handler   = sched_proc_update_handler,
 309                 .extra1         = &min_sched_tunable_scaling,
 310                 .extra2         = &max_sched_tunable_scaling,
 311         },
 312         {
 313                 .procname       = "sched_migration_cost",
 314                 .data           = &sysctl_sched_migration_cost,
 315                 .maxlen         = sizeof(unsigned int),
 316                 .mode           = 0644,
 317                 .proc_handler   = proc_dointvec,
 318         },
 319         {
 320                 .procname       = "sched_nr_migrate",
 321                 .data           = &sysctl_sched_nr_migrate,
 322                 .maxlen         = sizeof(unsigned int),
 323                 .mode           = 0644,
 324                 .proc_handler   = proc_dointvec,
 325         },
 326         {
 327                 .procname       = "sched_time_avg",
 328                 .data           = &sysctl_sched_time_avg,
 329                 .maxlen         = sizeof(unsigned int),
 330                 .mode           = 0644,
 331                 .proc_handler   = proc_dointvec,
 332         },
 333         {
 334                 .procname       = "sched_shares_window",
 335                 .data           = &sysctl_sched_shares_window,
 336                 .maxlen         = sizeof(unsigned int),
 337                 .mode           = 0644,
 338                 .proc_handler   = proc_dointvec,
 339         },
 340         {
 341                 .procname       = "timer_migration",
 342                 .data           = &sysctl_timer_migration,
 343                 .maxlen         = sizeof(unsigned int),
 344                 .mode           = 0644,
 345                 .proc_handler   = proc_dointvec_minmax,
 346                 .extra1         = &zero,
 347                 .extra2         = &one,
 348         },
 349 #endif
 350         {
 351                 .procname       = "sched_rt_period_us",
 352                 .data           = &sysctl_sched_rt_period,
 353                 .maxlen         = sizeof(unsigned int),
 354                 .mode           = 0644,
 355                 .proc_handler   = sched_rt_handler,
 356         },
 357         {
 358                 .procname       = "sched_rt_runtime_us",
 359                 .data           = &sysctl_sched_rt_runtime,
 360                 .maxlen         = sizeof(int),
 361                 .mode           = 0644,
 362                 .proc_handler   = sched_rt_handler,
 363         },
 364         {
 365                 .procname       = "sched_compat_yield",
 366                 .data           = &sysctl_sched_compat_yield,
 367                 .maxlen         = sizeof(unsigned int),
 368                 .mode           = 0644,
 369                 .proc_handler   = proc_dointvec,
 370         },
 371 #ifdef CONFIG_SCHED_AUTOGROUP
 372         {
 373                 .procname       = "sched_autogroup_enabled",
 374                 .data           = &sysctl_sched_autogroup_enabled,
 375                 .maxlen         = sizeof(unsigned int),
 376                 .mode           = 0644,
 377                 .proc_handler   = proc_dointvec,
 378                 .extra1         = &zero,
 379                 .extra2         = &one,
 380         },
 381 #endif
 382 #ifdef CONFIG_PROVE_LOCKING
 383         {
 384                 .procname       = "prove_locking",
 385                 .data           = &prove_locking,
 386                 .maxlen         = sizeof(int),
 387                 .mode           = 0644,
 388                 .proc_handler   = proc_dointvec,
 389         },
 390 #endif
 391 #ifdef CONFIG_LOCK_STAT
 392         {
 393                 .procname       = "lock_stat",
 394                 .data           = &lock_stat,
 395                 .maxlen         = sizeof(int),
 396                 .mode           = 0644,
 397                 .proc_handler   = proc_dointvec,
 398         },
 399 #endif
 400         {
 401                 .procname       = "panic",
 402                 .data           = &panic_timeout,
 403                 .maxlen         = sizeof(int),
 404                 .mode           = 0644,
 405                 .proc_handler   = proc_dointvec,
 406         },
 407         {
 408                 .procname       = "core_uses_pid",
 409                 .data           = &core_uses_pid,
 410                 .maxlen         = sizeof(int),
 411                 .mode           = 0644,
 412                 .proc_handler   = proc_dointvec,
 413         },
 414         {
 415                 .procname       = "core_pattern",
 416                 .data           = core_pattern,
 417                 .maxlen         = CORENAME_MAX_SIZE,
 418                 .mode           = 0644,
 419                 .proc_handler   = proc_dostring,
 420         },
 421         {
 422                 .procname       = "core_pipe_limit",
 423                 .data           = &core_pipe_limit,
 424                 .maxlen         = sizeof(unsigned int),
 425                 .mode           = 0644,
 426                 .proc_handler   = proc_dointvec,
 427         },
 428 #ifdef CONFIG_PROC_SYSCTL
 429         {
 430                 .procname       = "tainted",
 431                 .maxlen         = sizeof(long),
 432                 .mode           = 0644,
 433                 .proc_handler   = proc_taint,
 434         },
 435 #endif
 436 #ifdef CONFIG_LATENCYTOP
 437         {
 438                 .procname       = "latencytop",
 439                 .data           = &latencytop_enabled,
 440                 .maxlen         = sizeof(int),
 441                 .mode           = 0644,
 442                 .proc_handler   = proc_dointvec,
 443         },
 444 #endif
 445 #ifdef CONFIG_BLK_DEV_INITRD
 446         {
 447                 .procname       = "real-root-dev",
 448                 .data           = &real_root_dev,
 449                 .maxlen         = sizeof(int),
 450                 .mode           = 0644,
 451                 .proc_handler   = proc_dointvec,
 452         },
 453 #endif
 454         {
 455                 .procname       = "print-fatal-signals",
 456                 .data           = &print_fatal_signals,
 457                 .maxlen         = sizeof(int),
 458                 .mode           = 0644,
 459                 .proc_handler   = proc_dointvec,
 460         },
 461 #ifdef CONFIG_SPARC
 462         {
 463                 .procname       = "reboot-cmd",
 464                 .data           = reboot_command,
 465                 .maxlen         = 256,
 466                 .mode           = 0644,
 467                 .proc_handler   = proc_dostring,
 468         },
 469         {
 470                 .procname       = "stop-a",
 471                 .data           = &stop_a_enabled,
 472                 .maxlen         = sizeof (int),
 473                 .mode           = 0644,
 474                 .proc_handler   = proc_dointvec,
 475         },
 476         {
 477                 .procname       = "scons-poweroff",
 478                 .data           = &scons_pwroff,
 479                 .maxlen         = sizeof (int),
 480                 .mode           = 0644,
 481                 .proc_handler   = proc_dointvec,
 482         },
 483 #endif
 484 #ifdef CONFIG_SPARC64
 485         {
 486                 .procname       = "tsb-ratio",
 487                 .data           = &sysctl_tsb_ratio,
 488                 .maxlen         = sizeof (int),
 489                 .mode           = 0644,
 490                 .proc_handler   = proc_dointvec,
 491         },
 492 #endif
 493 #ifdef __hppa__
 494         {
 495                 .procname       = "soft-power",
 496                 .data           = &pwrsw_enabled,
 497                 .maxlen         = sizeof (int),
 498                 .mode           = 0644,
 499                 .proc_handler   = proc_dointvec,
 500         },
 501         {
 502                 .procname       = "unaligned-trap",
 503                 .data           = &unaligned_enabled,
 504                 .maxlen         = sizeof (int),
 505                 .mode           = 0644,
 506                 .proc_handler   = proc_dointvec,
 507         },
 508 #endif
 509         {
 510                 .procname       = "ctrl-alt-del",
 511                 .data           = &C_A_D,
 512                 .maxlen         = sizeof(int),
 513                 .mode           = 0644,
 514                 .proc_handler   = proc_dointvec,
 515         },
 516 #ifdef CONFIG_FUNCTION_TRACER
 517         {
 518                 .procname       = "ftrace_enabled",
 519                 .data           = &ftrace_enabled,
 520                 .maxlen         = sizeof(int),
 521                 .mode           = 0644,
 522                 .proc_handler   = ftrace_enable_sysctl,
 523         },
 524 #endif
 525 #ifdef CONFIG_STACK_TRACER
 526         {
 527                 .procname       = "stack_tracer_enabled",
 528                 .data           = &stack_tracer_enabled,
 529                 .maxlen         = sizeof(int),
 530                 .mode           = 0644,
 531                 .proc_handler   = stack_trace_sysctl,
 532         },
 533 #endif
 534 #ifdef CONFIG_TRACING
 535         {
 536                 .procname       = "ftrace_dump_on_oops",
 537                 .data           = &ftrace_dump_on_oops,
 538                 .maxlen         = sizeof(int),
 539                 .mode           = 0644,
 540                 .proc_handler   = proc_dointvec,
 541         },
 542 #endif
 543 #ifdef CONFIG_MODULES
 544         {
 545                 .procname       = "modprobe",
 546                 .data           = &modprobe_path,
 547                 .maxlen         = KMOD_PATH_LEN,
 548                 .mode           = 0644,
 549                 .proc_handler   = proc_dostring,
 550         },
 551         {
 552                 .procname       = "modules_disabled",
 553                 .data           = &modules_disabled,
 554                 .maxlen         = sizeof(int),
 555                 .mode           = 0644,
 556                 /* only handle a transition from default "0" to "1" */
 557                 .proc_handler   = proc_dointvec_minmax,
 558                 .extra1         = &one,
 559                 .extra2         = &one,
 560         },
 561 #endif
 562 #ifdef CONFIG_HOTPLUG
 563         {
 564                 .procname       = "hotplug",
 565                 .data           = &uevent_helper,
 566                 .maxlen         = UEVENT_HELPER_PATH_LEN,
 567                 .mode           = 0644,
 568                 .proc_handler   = proc_dostring,
 569         },
 570 #endif
 571 #ifdef CONFIG_CHR_DEV_SG
 572         {
 573                 .procname       = "sg-big-buff",
 574                 .data           = &sg_big_buff,
 575                 .maxlen         = sizeof (int),
 576                 .mode           = 0444,
 577                 .proc_handler   = proc_dointvec,
 578         },
 579 #endif
 580 #ifdef CONFIG_BSD_PROCESS_ACCT
 581         {
 582                 .procname       = "acct",
 583                 .data           = &acct_parm,
 584                 .maxlen         = 3*sizeof(int),
 585                 .mode           = 0644,
 586                 .proc_handler   = proc_dointvec,
 587         },
 588 #endif
 589 #ifdef CONFIG_MAGIC_SYSRQ
 590         {
 591                 .procname       = "sysrq",
 592                 .data           = &__sysrq_enabled,
 593                 .maxlen         = sizeof (int),
 594                 .mode           = 0644,
 595                 .proc_handler   = sysrq_sysctl_handler,
 596         },
 597 #endif
 598 #ifdef CONFIG_PROC_SYSCTL
 599         {
 600                 .procname       = "cad_pid",
 601                 .data           = NULL,
 602                 .maxlen         = sizeof (int),
 603                 .mode           = 0600,
 604                 .proc_handler   = proc_do_cad_pid,
 605         },
 606 #endif
 607         {
 608                 .procname       = "threads-max",
 609                 .data           = &max_threads,
 610                 .maxlen         = sizeof(int),
 611                 .mode           = 0644,
 612                 .proc_handler   = proc_dointvec,
 613         },
 614         {
 615                 .procname       = "random",
 616                 .mode           = 0555,
 617                 .child          = random_table,
 618         },
 619         {
 620                 .procname       = "overflowuid",
 621                 .data           = &overflowuid,
 622                 .maxlen         = sizeof(int),
 623                 .mode           = 0644,
 624                 .proc_handler   = proc_dointvec_minmax,
 625                 .extra1         = &minolduid,
 626                 .extra2         = &maxolduid,
 627         },
 628         {
 629                 .procname       = "overflowgid",
 630                 .data           = &overflowgid,
 631                 .maxlen         = sizeof(int),
 632                 .mode           = 0644,
 633                 .proc_handler   = proc_dointvec_minmax,
 634                 .extra1         = &minolduid,
 635                 .extra2         = &maxolduid,
 636         },
 637 #ifdef CONFIG_S390
 638 #ifdef CONFIG_MATHEMU
 639         {
 640                 .procname       = "ieee_emulation_warnings",
 641                 .data           = &sysctl_ieee_emulation_warnings,
 642                 .maxlen         = sizeof(int),
 643                 .mode           = 0644,
 644                 .proc_handler   = proc_dointvec,
 645         },
 646 #endif
 647         {
 648                 .procname       = "userprocess_debug",
 649                 .data           = &show_unhandled_signals,
 650                 .maxlen         = sizeof(int),
 651                 .mode           = 0644,
 652                 .proc_handler   = proc_dointvec,
 653         },
 654 #endif
 655         {
 656                 .procname       = "pid_max",
 657                 .data           = &pid_max,
 658                 .maxlen         = sizeof (int),
 659                 .mode           = 0644,
 660                 .proc_handler   = proc_dointvec_minmax,
 661                 .extra1         = &pid_max_min,
 662                 .extra2         = &pid_max_max,
 663         },
 664         {
 665                 .procname       = "panic_on_oops",
 666                 .data           = &panic_on_oops,
 667                 .maxlen         = sizeof(int),
 668                 .mode           = 0644,
 669                 .proc_handler   = proc_dointvec,
 670         },
 671 #if defined CONFIG_PRINTK
 672         {
 673                 .procname       = "printk",
 674                 .data           = &console_loglevel,
 675                 .maxlen         = 4*sizeof(int),
 676                 .mode           = 0644,
 677                 .proc_handler   = proc_dointvec,
 678         },
 679         {
 680                 .procname       = "printk_ratelimit",
 681                 .data           = &printk_ratelimit_state.interval,
 682                 .maxlen         = sizeof(int),
 683                 .mode           = 0644,
 684                 .proc_handler   = proc_dointvec_jiffies,
 685         },
 686         {
 687                 .procname       = "printk_ratelimit_burst",
 688                 .data           = &printk_ratelimit_state.burst,
 689                 .maxlen         = sizeof(int),
 690                 .mode           = 0644,
 691                 .proc_handler   = proc_dointvec,
 692         },
 693         {
 694                 .procname       = "printk_delay",
 695                 .data           = &printk_delay_msec,
 696                 .maxlen         = sizeof(int),
 697                 .mode           = 0644,
 698                 .proc_handler   = proc_dointvec_minmax,
 699                 .extra1         = &zero,
 700                 .extra2         = &ten_thousand,
 701         },
 702         {
 703                 .procname       = "dmesg_restrict",
 704                 .data           = &dmesg_restrict,
 705                 .maxlen         = sizeof(int),
 706                 .mode           = 0644,
 707                 .proc_handler   = proc_dointvec_minmax,
 708                 .extra1         = &zero,
 709                 .extra2         = &one,
 710         },
 711         {
 712                 .procname       = "kptr_restrict",
 713                 .data           = &kptr_restrict,
 714                 .maxlen         = sizeof(int),
 715                 .mode           = 0644,
 716                 .proc_handler   = proc_dointvec_minmax,
 717                 .extra1         = &zero,
 718                 .extra2         = &two,
 719         },
 720 #endif
 721         {
 722                 .procname       = "ngroups_max",
 723                 .data           = &ngroups_max,
 724                 .maxlen         = sizeof (int),
 725                 .mode           = 0444,
 726                 .proc_handler   = proc_dointvec,
 727         },
 728 #if defined(CONFIG_LOCKUP_DETECTOR)
 729         {
 730                 .procname       = "watchdog",
 731                 .data           = &watchdog_enabled,
 732                 .maxlen         = sizeof (int),
 733                 .mode           = 0644,
 734                 .proc_handler   = proc_dowatchdog_enabled,
 735         },
 736         {
 737                 .procname       = "watchdog_thresh",
 738                 .data           = &softlockup_thresh,
 739                 .maxlen         = sizeof(int),
 740                 .mode           = 0644,
 741                 .proc_handler   = proc_dowatchdog_thresh,
 742                 .extra1         = &neg_one,
 743                 .extra2         = &sixty,
 744         },
 745         {
 746                 .procname       = "softlockup_panic",
 747                 .data           = &softlockup_panic,
 748                 .maxlen         = sizeof(int),
 749                 .mode           = 0644,
 750                 .proc_handler   = proc_dointvec_minmax,
 751                 .extra1         = &zero,
 752                 .extra2         = &one,
 753         },
 754         {
 755                 .procname       = "nmi_watchdog",
 756                 .data           = &watchdog_enabled,
 757                 .maxlen         = sizeof (int),
 758                 .mode           = 0644,
 759                 .proc_handler   = proc_dowatchdog_enabled,
 760         },
 761 #endif
 762 #if defined(CONFIG_X86_LOCAL_APIC) && defined(CONFIG_X86)
 763         {
 764                 .procname       = "unknown_nmi_panic",
 765                 .data           = &unknown_nmi_panic,
 766                 .maxlen         = sizeof (int),
 767                 .mode           = 0644,
 768                 .proc_handler   = proc_dointvec,
 769         },
 770 #endif
 771 #if defined(CONFIG_X86)
 772         {
 773                 .procname       = "panic_on_unrecovered_nmi",
 774                 .data           = &panic_on_unrecovered_nmi,
 775                 .maxlen         = sizeof(int),
 776                 .mode           = 0644,
 777                 .proc_handler   = proc_dointvec,
 778         },
 779         {
 780                 .procname       = "panic_on_io_nmi",
 781                 .data           = &panic_on_io_nmi,
 782                 .maxlen         = sizeof(int),
 783                 .mode           = 0644,
 784                 .proc_handler   = proc_dointvec,
 785         },
 786         {
 787                 .procname       = "bootloader_type",
 788                 .data           = &bootloader_type,
 789                 .maxlen         = sizeof (int),
 790                 .mode           = 0444,
 791                 .proc_handler   = proc_dointvec,
 792         },
 793         {
 794                 .procname       = "bootloader_version",
 795                 .data           = &bootloader_version,
 796                 .maxlen         = sizeof (int),
 797                 .mode           = 0444,
 798                 .proc_handler   = proc_dointvec,
 799         },
 800         {
 801                 .procname       = "kstack_depth_to_print",
 802                 .data           = &kstack_depth_to_print,
 803                 .maxlen         = sizeof(int),
 804                 .mode           = 0644,
 805                 .proc_handler   = proc_dointvec,
 806         },
 807         {
 808                 .procname       = "io_delay_type",
 809                 .data           = &io_delay_type,
 810                 .maxlen         = sizeof(int),
 811                 .mode           = 0644,
 812                 .proc_handler   = proc_dointvec,
 813         },
 814 #endif
 815 #if defined(CONFIG_MMU)
 816         {
 817                 .procname       = "randomize_va_space",
 818                 .data           = &randomize_va_space,
 819                 .maxlen         = sizeof(int),
 820                 .mode           = 0644,
 821                 .proc_handler   = proc_dointvec,
 822         },
 823 #endif
 824 #if defined(CONFIG_S390) && defined(CONFIG_SMP)
 825         {
 826                 .procname       = "spin_retry",
 827                 .data           = &spin_retry,
 828                 .maxlen         = sizeof (int),
 829                 .mode           = 0644,
 830                 .proc_handler   = proc_dointvec,
 831         },
 832 #endif
 833 #if     defined(CONFIG_ACPI_SLEEP) && defined(CONFIG_X86)
 834         {
 835                 .procname       = "acpi_video_flags",
 836                 .data           = &acpi_realmode_flags,
 837                 .maxlen         = sizeof (unsigned long),
 838                 .mode           = 0644,
 839                 .proc_handler   = proc_doulongvec_minmax,
 840         },
 841 #endif
 842 #ifdef CONFIG_IA64
 843         {
 844                 .procname       = "ignore-unaligned-usertrap",
 845                 .data           = &no_unaligned_warning,
 846                 .maxlen         = sizeof (int),
 847                 .mode           = 0644,
 848                 .proc_handler   = proc_dointvec,
 849         },
 850         {
 851                 .procname       = "unaligned-dump-stack",
 852                 .data           = &unaligned_dump_stack,
 853                 .maxlen         = sizeof (int),
 854                 .mode           = 0644,
 855                 .proc_handler   = proc_dointvec,
 856         },
 857 #endif
 858 #ifdef CONFIG_DETECT_HUNG_TASK
 859         {
 860                 .procname       = "hung_task_panic",
 861                 .data           = &sysctl_hung_task_panic,
 862                 .maxlen         = sizeof(int),
 863                 .mode           = 0644,
 864                 .proc_handler   = proc_dointvec_minmax,
 865                 .extra1         = &zero,
 866                 .extra2         = &one,
 867         },
 868         {
 869                 .procname       = "hung_task_check_count",
 870                 .data           = &sysctl_hung_task_check_count,
 871                 .maxlen         = sizeof(unsigned long),
 872                 .mode           = 0644,
 873                 .proc_handler   = proc_doulongvec_minmax,
 874         },
 875         {
 876                 .procname       = "hung_task_timeout_secs",
 877                 .data           = &sysctl_hung_task_timeout_secs,
 878                 .maxlen         = sizeof(unsigned long),
 879                 .mode           = 0644,
 880                 .proc_handler   = proc_dohung_task_timeout_secs,
 881         },
 882         {
 883                 .procname       = "hung_task_warnings",
 884                 .data           = &sysctl_hung_task_warnings,
 885                 .maxlen         = sizeof(unsigned long),
 886                 .mode           = 0644,
 887                 .proc_handler   = proc_doulongvec_minmax,
 888         },
 889 #endif
 890 #ifdef CONFIG_COMPAT
 891         {
 892                 .procname       = "compat-log",
 893                 .data           = &compat_log,
 894                 .maxlen         = sizeof (int),
 895                 .mode           = 0644,
 896                 .proc_handler   = proc_dointvec,
 897         },
 898 #endif
 899 #ifdef CONFIG_RT_MUTEXES
 900         {
 901                 .procname       = "max_lock_depth",
 902                 .data           = &max_lock_depth,
 903                 .maxlen         = sizeof(int),
 904                 .mode           = 0644,
 905                 .proc_handler   = proc_dointvec,
 906         },
 907 #endif
 908         {
 909                 .procname       = "poweroff_cmd",
 910                 .data           = &poweroff_cmd,
 911                 .maxlen         = POWEROFF_CMD_PATH_LEN,
 912                 .mode           = 0644,
 913                 .proc_handler   = proc_dostring,
 914         },
 915 #ifdef CONFIG_KEYS
 916         {
 917                 .procname       = "keys",
 918                 .mode           = 0555,
 919                 .child          = key_sysctls,
 920         },
 921 #endif
 922 #ifdef CONFIG_RCU_TORTURE_TEST
 923         {
 924                 .procname       = "rcutorture_runnable",
 925                 .data           = &rcutorture_runnable,
 926                 .maxlen         = sizeof(int),
 927                 .mode           = 0644,
 928                 .proc_handler   = proc_dointvec,
 929         },
 930 #endif
 931 #ifdef CONFIG_PERF_EVENTS
 932         {
 933                 .procname       = "perf_event_paranoid",
 934                 .data           = &sysctl_perf_event_paranoid,
 935                 .maxlen         = sizeof(sysctl_perf_event_paranoid),
 936                 .mode           = 0644,
 937                 .proc_handler   = proc_dointvec,
 938         },
 939         {
 940                 .procname       = "perf_event_mlock_kb",
 941                 .data           = &sysctl_perf_event_mlock,
 942                 .maxlen         = sizeof(sysctl_perf_event_mlock),
 943                 .mode           = 0644,
 944                 .proc_handler   = proc_dointvec,
 945         },
 946         {
 947                 .procname       = "perf_event_max_sample_rate",
 948                 .data           = &sysctl_perf_event_sample_rate,
 949                 .maxlen         = sizeof(sysctl_perf_event_sample_rate),
 950                 .mode           = 0644,
 951                 .proc_handler   = proc_dointvec,
 952         },
 953 #endif
 954 #ifdef CONFIG_KMEMCHECK
 955         {
 956                 .procname       = "kmemcheck",
 957                 .data           = &kmemcheck_enabled,
 958                 .maxlen         = sizeof(int),
 959                 .mode           = 0644,
 960                 .proc_handler   = proc_dointvec,
 961         },
 962 #endif
 963 #ifdef CONFIG_BLOCK
 964         {
 965                 .procname       = "blk_iopoll",
 966                 .data           = &blk_iopoll_enabled,
 967                 .maxlen         = sizeof(int),
 968                 .mode           = 0644,
 969                 .proc_handler   = proc_dointvec,
 970         },
 971 #endif
 972         { }
 973 };
 974
 975 static struct ctl_table vm_table[] = {
 976         {
 977                 .procname       = "overcommit_memory",
 978                 .data           = &sysctl_overcommit_memory,
 979                 .maxlen         = sizeof(sysctl_overcommit_memory),
 980                 .mode           = 0644,
 981                 .proc_handler   = proc_dointvec,
 982         },
 983         {
 984                 .procname       = "panic_on_oom",
 985                 .data           = &sysctl_panic_on_oom,
 986                 .maxlen         = sizeof(sysctl_panic_on_oom),
 987                 .mode           = 0644,
 988                 .proc_handler   = proc_dointvec,
 989         },
 990         {
 991                 .procname       = "oom_kill_allocating_task",
 992                 .data           = &sysctl_oom_kill_allocating_task,
 993                 .maxlen         = sizeof(sysctl_oom_kill_allocating_task),
 994                 .mode           = 0644,
 995                 .proc_handler   = proc_dointvec,
 996         },
 997         {
 998                 .procname       = "oom_dump_tasks",
 999                 .data           = &sysctl_oom_dump_tasks,
1000                 .maxlen         = sizeof(sysctl_oom_dump_tasks),
1001                 .mode           = 0644,
1002                 .proc_handler   = proc_dointvec,
1003         },
1004         {
1005                 .procname       = "overcommit_ratio",
1006                 .data           = &sysctl_overcommit_ratio,
1007                 .maxlen         = sizeof(sysctl_overcommit_ratio),
1008                 .mode           = 0644,
1009                 .proc_handler   = proc_dointvec,
1010         },
1011         {
1012                 .procname       = "page-cluster",
1013                 .data           = &page_cluster,
1014                 .maxlen         = sizeof(int),
1015                 .mode           = 0644,
1016                 .proc_handler   = proc_dointvec,
1017         },
1018         {
1019                 .procname       = "dirty_background_ratio",
1020                 .data           = &dirty_background_ratio,
1021                 .maxlen         = sizeof(dirty_background_ratio),
1022                 .mode           = 0644,
1023                 .proc_handler   = dirty_background_ratio_handler,
1024                 .extra1         = &zero,
1025                 .extra2         = &one_hundred,
1026         },
1027         {
1028                 .procname       = "dirty_background_bytes",
1029                 .data           = &dirty_background_bytes,
1030                 .maxlen         = sizeof(dirty_background_bytes),
1031                 .mode           = 0644,
1032                 .proc_handler   = dirty_background_bytes_handler,
1033                 .extra1         = &one_ul,
1034         },
1035         {
1036                 .procname       = "dirty_ratio",
1037                 .data           = &vm_dirty_ratio,
1038                 .maxlen         = sizeof(vm_dirty_ratio),
1039                 .mode           = 0644,
1040                 .proc_handler   = dirty_ratio_handler,
1041                 .extra1         = &zero,
1042                 .extra2         = &one_hundred,
1043         },
1044         {
1045                 .procname       = "dirty_bytes",
1046                 .data           = &vm_dirty_bytes,
1047                 .maxlen         = sizeof(vm_dirty_bytes),
1048                 .mode           = 0644,
1049                 .proc_handler   = dirty_bytes_handler,
1050                 .extra1         = &dirty_bytes_min,
1051         },
1052         {
1053                 .procname       = "dirty_writeback_centisecs",
1054                 .data           = &dirty_writeback_interval,
1055                 .maxlen         = sizeof(dirty_writeback_interval),
1056                 .mode           = 0644,
1057                 .proc_handler   = dirty_writeback_centisecs_handler,
1058         },
1059         {
1060                 .procname       = "dirty_expire_centisecs",
1061                 .data           = &dirty_expire_interval,
1062                 .maxlen         = sizeof(dirty_expire_interval),
1063                 .mode           = 0644,
1064                 .proc_handler   = proc_dointvec,
1065         },
1066         {
1067                 .procname       = "nr_pdflush_threads",
1068                 .data           = &nr_pdflush_threads,
1069                 .maxlen         = sizeof nr_pdflush_threads,
1070                 .mode           = 0444 /* read-only*/,
1071                 .proc_handler   = proc_dointvec,
1072         },
1073         {
1074                 .procname       = "swappiness",
1075                 .data           = &vm_swappiness,
1076                 .maxlen         = sizeof(vm_swappiness),
1077                 .mode           = 0644,
1078                 .proc_handler   = proc_dointvec_minmax,
1079                 .extra1         = &zero,
1080                 .extra2         = &one_hundred,
1081         },
1082 #ifdef CONFIG_HUGETLB_PAGE
1083         {
1084                 .procname       = "nr_hugepages",
1085                 .data           = NULL,
1086                 .maxlen         = sizeof(unsigned long),
1087                 .mode           = 0644,
1088                 .proc_handler   = hugetlb_sysctl_handler,
1089                 .extra1         = (void *)&hugetlb_zero,
1090                 .extra2         = (void *)&hugetlb_infinity,
1091         },
1092 #ifdef CONFIG_NUMA
1093         {
1094                 .procname       = "nr_hugepages_mempolicy",
1095                 .data           = NULL,
1096                 .maxlen         = sizeof(unsigned long),
1097                 .mode           = 0644,
1098                 .proc_handler   = &hugetlb_mempolicy_sysctl_handler,
1099                 .extra1         = (void *)&hugetlb_zero,
1100                 .extra2         = (void *)&hugetlb_infinity,
1101         },
1102 #endif
1103          {
1104                 .procname       = "hugetlb_shm_group",
1105                 .data           = &sysctl_hugetlb_shm_group,
1106                 .maxlen         = sizeof(gid_t),
1107                 .mode           = 0644,
1108                 .proc_handler   = proc_dointvec,
1109          },
1110          {
1111                 .procname       = "hugepages_treat_as_movable",
1112                 .data           = &hugepages_treat_as_movable,
1113                 .maxlen         = sizeof(int),
1114                 .mode           = 0644,
1115                 .proc_handler   = hugetlb_treat_movable_handler,
1116         },
1117         {
1118                 .procname       = "nr_overcommit_hugepages",
1119                 .data           = NULL,
1120                 .maxlen         = sizeof(unsigned long),
1121                 .mode           = 0644,
1122                 .proc_handler   = hugetlb_overcommit_handler,
1123                 .extra1         = (void *)&hugetlb_zero,
1124                 .extra2         = (void *)&hugetlb_infinity,
1125         },
1126 #endif
1127         {
1128                 .procname       = "lowmem_reserve_ratio",
1129                 .data           = &sysctl_lowmem_reserve_ratio,
1130                 .maxlen         = sizeof(sysctl_lowmem_reserve_ratio),
1131                 .mode           = 0644,
1132                 .proc_handler   = lowmem_reserve_ratio_sysctl_handler,
1133         },
1134         {
1135                 .procname       = "drop_caches",
1136                 .data           = &sysctl_drop_caches,
1137                 .maxlen         = sizeof(int),
1138                 .mode           = 0644,
1139                 .proc_handler   = drop_caches_sysctl_handler,
1140         },
1141 #ifdef CONFIG_COMPACTION
1142         {
1143                 .procname       = "compact_memory",
1144                 .data           = &sysctl_compact_memory,
1145                 .maxlen         = sizeof(int),
1146                 .mode           = 0200,
1147                 .proc_handler   = sysctl_compaction_handler,
1148         },
1149         {
1150                 .procname       = "extfrag_threshold",
1151                 .data           = &sysctl_extfrag_threshold,
1152                 .maxlen         = sizeof(int),
1153                 .mode           = 0644,
1154                 .proc_handler   = sysctl_extfrag_handler,
1155                 .extra1         = &min_extfrag_threshold,
1156                 .extra2         = &max_extfrag_threshold,
1157         },
1158
1159 #endif /* CONFIG_COMPACTION */
1160         {
1161                 .procname       = "min_free_kbytes",
1162                 .data           = &min_free_kbytes,
1163                 .maxlen         = sizeof(min_free_kbytes),
1164                 .mode           = 0644,
1165                 .proc_handler   = min_free_kbytes_sysctl_handler,
1166                 .extra1         = &zero,
1167         },
1168         {
1169                 .procname       = "percpu_pagelist_fraction",
1170                 .data           = &percpu_pagelist_fraction,
1171                 .maxlen         = sizeof(percpu_pagelist_fraction),
1172                 .mode           = 0644,
1173                 .proc_handler   = percpu_pagelist_fraction_sysctl_handler,
1174                 .extra1         = &min_percpu_pagelist_fract,
1175         },
1176 #ifdef CONFIG_MMU
1177         {
1178                 .procname       = "max_map_count",
1179                 .data           = &sysctl_max_map_count,
1180                 .maxlen         = sizeof(sysctl_max_map_count),
1181                 .mode           = 0644,
1182                 .proc_handler   = proc_dointvec_minmax,
1183                 .extra1         = &zero,
1184         },
1185 #else
1186         {
1187                 .procname       = "nr_trim_pages",
1188                 .data           = &sysctl_nr_trim_pages,
1189                 .maxlen         = sizeof(sysctl_nr_trim_pages),
1190                 .mode           = 0644,
1191                 .proc_handler   = proc_dointvec_minmax,
1192                 .extra1         = &zero,
1193         },
1194 #endif
1195         {
1196                 .procname       = "laptop_mode",
1197                 .data           = &laptop_mode,
1198                 .maxlen         = sizeof(laptop_mode),
1199                 .mode           = 0644,
1200                 .proc_handler   = proc_dointvec_jiffies,
1201         },
1202         {
1203                 .procname       = "block_dump",
1204                 .data           = &block_dump,
1205                 .maxlen         = sizeof(block_dump),
1206                 .mode           = 0644,
1207                 .proc_handler   = proc_dointvec,
1208                 .extra1         = &zero,
1209         },
1210         {
1211                 .procname       = "vfs_cache_pressure",
1212                 .data           = &sysctl_vfs_cache_pressure,
1213                 .maxlen         = sizeof(sysctl_vfs_cache_pressure),
1214                 .mode           = 0644,
1215                 .proc_handler   = proc_dointvec,
1216                 .extra1         = &zero,
1217         },
1218 #ifdef HAVE_ARCH_PICK_MMAP_LAYOUT
1219         {
1220                 .procname       = "legacy_va_layout",
1221                 .data           = &sysctl_legacy_va_layout,
1222                 .maxlen         = sizeof(sysctl_legacy_va_layout),
1223                 .mode           = 0644,
1224                 .proc_handler   = proc_dointvec,
1225                 .extra1         = &zero,
1226         },
1227 #endif
1228 #ifdef CONFIG_NUMA
1229         {
1230                 .procname       = "zone_reclaim_mode",
1231                 .data           = &zone_reclaim_mode,
1232                 .maxlen         = sizeof(zone_reclaim_mode),
1233                 .mode           = 0644,
1234                 .proc_handler   = proc_dointvec,
1235                 .extra1         = &zero,
1236         },
1237         {
1238                 .procname       = "min_unmapped_ratio",
1239                 .data           = &sysctl_min_unmapped_ratio,
1240                 .maxlen         = sizeof(sysctl_min_unmapped_ratio),
1241                 .mode           = 0644,
1242                 .proc_handler   = sysctl_min_unmapped_ratio_sysctl_handler,
1243                 .extra1         = &zero,
1244                 .extra2         = &one_hundred,
1245         },
1246         {
1247                 .procname       = "min_slab_ratio",
1248                 .data           = &sysctl_min_slab_ratio,
1249                 .maxlen         = sizeof(sysctl_min_slab_ratio),
1250                 .mode           = 0644,
1251                 .proc_handler   = sysctl_min_slab_ratio_sysctl_handler,
1252                 .extra1         = &zero,
1253                 .extra2         = &one_hundred,
1254         },
1255 #endif
1256 #ifdef CONFIG_SMP
1257         {
1258                 .procname       = "stat_interval",
1259                 .data           = &sysctl_stat_interval,
1260                 .maxlen         = sizeof(sysctl_stat_interval),
1261                 .mode           = 0644,
1262                 .proc_handler   = proc_dointvec_jiffies,
1263         },
1264 #endif
1265 #ifdef CONFIG_MMU
1266         {
1267                 .procname       = "mmap_min_addr",
1268                 .data           = &dac_mmap_min_addr,
1269                 .maxlen         = sizeof(unsigned long),
1270                 .mode           = 0644,
1271                 .proc_handler   = mmap_min_addr_handler,
1272         },
1273 #endif
1274 #ifdef CONFIG_NUMA
1275         {
1276                 .procname       = "numa_zonelist_order",
1277                 .data           = &numa_zonelist_order,
1278                 .maxlen         = NUMA_ZONELIST_ORDER_LEN,
1279                 .mode           = 0644,
1280                 .proc_handler   = numa_zonelist_order_handler,
1281         },
1282 #endif
1283 #if (defined(CONFIG_X86_32) && !defined(CONFIG_UML))|| \
1284    (defined(CONFIG_SUPERH) && defined(CONFIG_VSYSCALL))
1285         {
1286                 .procname       = "vdso_enabled",
1287                 .data           = &vdso_enabled,
1288                 .maxlen         = sizeof(vdso_enabled),
1289                 .mode           = 0644,
1290                 .proc_handler   = proc_dointvec,
1291                 .extra1         = &zero,
1292         },
1293 #endif
1294 #ifdef CONFIG_HIGHMEM
1295         {
1296                 .procname       = "highmem_is_dirtyable",
1297                 .data           = &vm_highmem_is_dirtyable,
1298                 .maxlen         = sizeof(vm_highmem_is_dirtyable),
1299                 .mode           = 0644,
1300                 .proc_handler   = proc_dointvec_minmax,
1301                 .extra1         = &zero,
1302                 .extra2         = &one,
1303         },
1304 #endif
1305         {
1306                 .procname       = "scan_unevictable_pages",
1307                 .data           = &scan_unevictable_pages,
1308                 .maxlen         = sizeof(scan_unevictable_pages),
1309                 .mode           = 0644,
1310                 .proc_handler   = scan_unevictable_handler,
1311         },
1312 #ifdef CONFIG_MEMORY_FAILURE
1313         {
1314                 .procname       = "memory_failure_early_kill",
1315                 .data           = &sysctl_memory_failure_early_kill,
1316                 .maxlen         = sizeof(sysctl_memory_failure_early_kill),
1317                 .mode           = 0644,
1318                 .proc_handler   = proc_dointvec_minmax,
1319                 .extra1         = &zero,
1320                 .extra2         = &one,
1321         },
1322         {
1323                 .procname       = "memory_failure_recovery",
1324                 .data           = &sysctl_memory_failure_recovery,
1325                 .maxlen         = sizeof(sysctl_memory_failure_recovery),
1326                 .mode           = 0644,
1327                 .proc_handler   = proc_dointvec_minmax,
1328                 .extra1         = &zero,
1329                 .extra2         = &one,
1330         },
1331 #endif
1332         { }
1333 };
1334
1335 #if defined(CONFIG_BINFMT_MISC) || defined(CONFIG_BINFMT_MISC_MODULE)
1336 static struct ctl_table binfmt_misc_table[] = {
1337         { }
1338 };
1339 #endif
1340
1341 static struct ctl_table fs_table[] = {
1342         {
1343                 .procname       = "inode-nr",
1344                 .data           = &inodes_stat,
1345                 .maxlen         = 2*sizeof(int),
1346                 .mode           = 0444,
1347                 .proc_handler   = proc_nr_inodes,
1348         },
1349         {
1350                 .procname       = "inode-state",
1351                 .data           = &inodes_stat,
1352                 .maxlen         = 7*sizeof(int),
1353                 .mode           = 0444,
1354                 .proc_handler   = proc_nr_inodes,
1355         },
1356         {
1357                 .procname       = "file-nr",
1358                 .data           = &files_stat,
1359                 .maxlen         = sizeof(files_stat),
1360                 .mode           = 0444,
1361                 .proc_handler   = proc_nr_files,
1362         },
1363         {
1364                 .procname       = "file-max",
1365                 .data           = &files_stat.max_files,
1366                 .maxlen         = sizeof(files_stat.max_files),
1367                 .mode           = 0644,
1368                 .proc_handler   = proc_doulongvec_minmax,
1369         },
1370         {
1371                 .procname       = "nr_open",
1372                 .data           = &sysctl_nr_open,
1373                 .maxlen         = sizeof(int),
1374                 .mode           = 0644,
1375                 .proc_handler   = proc_dointvec_minmax,
1376                 .extra1         = &sysctl_nr_open_min,
1377                 .extra2         = &sysctl_nr_open_max,
1378         },
1379         {
1380                 .procname       = "dentry-state",
1381                 .data           = &dentry_stat,
1382                 .maxlen         = 6*sizeof(int),
1383                 .mode           = 0444,
1384                 .proc_handler   = proc_nr_dentry,
1385         },
1386         {
1387                 .procname       = "overflowuid",
1388                 .data           = &fs_overflowuid,
1389                 .maxlen         = sizeof(int),
1390                 .mode           = 0644,
1391                 .proc_handler   = proc_dointvec_minmax,
1392                 .extra1         = &minolduid,
1393                 .extra2         = &maxolduid,
1394         },
1395         {
1396                 .procname       = "overflowgid",
1397                 .data           = &fs_overflowgid,
1398                 .maxlen         = sizeof(int),
1399                 .mode           = 0644,
1400                 .proc_handler   = proc_dointvec_minmax,
1401                 .extra1         = &minolduid,
1402                 .extra2         = &maxolduid,
1403         },
1404 #ifdef CONFIG_FILE_LOCKING
1405         {
1406                 .procname       = "leases-enable",
1407                 .data           = &leases_enable,
1408                 .maxlen         = sizeof(int),
1409                 .mode           = 0644,
1410                 .proc_handler   = proc_dointvec,
1411         },
1412 #endif
1413 #ifdef CONFIG_DNOTIFY
1414         {
1415                 .procname       = "dir-notify-enable",
1416                 .data           = &dir_notify_enable,
1417                 .maxlen         = sizeof(int),
1418                 .mode           = 0644,
1419                 .proc_handler   = proc_dointvec,
1420         },
1421 #endif
1422 #ifdef CONFIG_MMU
1423 #ifdef CONFIG_FILE_LOCKING
1424         {
1425                 .procname       = "lease-break-time",
1426                 .data           = &lease_break_time,
1427                 .maxlen         = sizeof(int),
1428                 .mode           = 0644,
1429                 .proc_handler   = proc_dointvec,
1430         },
1431 #endif
1432 #ifdef CONFIG_AIO
1433         {
1434                 .procname       = "aio-nr",
1435                 .data           = &aio_nr,
1436                 .maxlen         = sizeof(aio_nr),
1437                 .mode           = 0444,
1438                 .proc_handler   = proc_doulongvec_minmax,
1439         },
1440         {
1441                 .procname       = "aio-max-nr",
1442                 .data           = &aio_max_nr,
1443                 .maxlen         = sizeof(aio_max_nr),
1444                 .mode           = 0644,
1445                 .proc_handler   = proc_doulongvec_minmax,
1446         },
1447 #endif /* CONFIG_AIO */
1448 #ifdef CONFIG_INOTIFY_USER
1449         {
1450                 .procname       = "inotify",
1451                 .mode           = 0555,
1452                 .child          = inotify_table,
1453         },
1454 #endif
1455 #ifdef CONFIG_EPOLL
1456         {
1457                 .procname       = "epoll",
1458                 .mode           = 0555,
1459                 .child          = epoll_table,
1460         },
1461 #endif
1462 #endif
1463         {
1464                 .procname       = "suid_dumpable",
1465                 .data           = &suid_dumpable,
1466                 .maxlen         = sizeof(int),
1467                 .mode           = 0644,
1468                 .proc_handler   = proc_dointvec_minmax,
1469                 .extra1         = &zero,
1470                 .extra2         = &two,
1471         },
1472 #if defined(CONFIG_BINFMT_MISC) || defined(CONFIG_BINFMT_MISC_MODULE)
1473         {
1474                 .procname       = "binfmt_misc",
1475                 .mode           = 0555,
1476                 .child          = binfmt_misc_table,
1477         },
1478 #endif
1479         {
1480                 .procname       = "pipe-max-size",
1481                 .data           = &pipe_max_size,
1482                 .maxlen         = sizeof(int),
1483                 .mode           = 0644,
1484                 .proc_handler   = &pipe_proc_fn,
1485                 .extra1         = &pipe_min_size,
1486         },
1487         { }
1488 };
1489
1490 static struct ctl_table debug_table[] = {
1491 #if defined(CONFIG_X86) || defined(CONFIG_PPC) || defined(CONFIG_SPARC) || \
1492     defined(CONFIG_S390)
1493         {
1494                 .procname       = "exception-trace",
1495                 .data           = &show_unhandled_signals,
1496                 .maxlen         = sizeof(int),
1497                 .mode           = 0644,
1498                 .proc_handler   = proc_dointvec
1499         },
1500 #endif
1501 #if defined(CONFIG_OPTPROBES)
1502         {
1503                 .procname       = "kprobes-optimization",
1504                 .data           = &sysctl_kprobes_optimization,
1505                 .maxlen         = sizeof(int),
1506                 .mode           = 0644,
1507                 .proc_handler   = proc_kprobes_optimization_handler,
1508                 .extra1         = &zero,
1509                 .extra2         = &one,
1510         },
1511 #endif
1512         { }
1513 };
1514
1515 static struct ctl_table dev_table[] = {
1516         { }
1517 };
1518
1519 static DEFINE_SPINLOCK(sysctl_lock);
1520
1521 /* called under sysctl_lock */
1522 static int use_table(struct ctl_table_header *p)
1523 {
1524         if (unlikely(p->unregistering))
1525                 return 0;
1526         p->used++;
1527         return 1;
1528 }
1529
1530 /* called under sysctl_lock */
1531 static void unuse_table(struct ctl_table_header *p)
1532 {
1533         if (!--p->used)
1534                 if (unlikely(p->unregistering))
1535                         complete(p->unregistering);
1536 }
1537
1538 /* called under sysctl_lock, will reacquire if has to wait */
1539 static void start_unregistering(struct ctl_table_header *p)
1540 {
1541         /*
1542          * if p->used is 0, nobody will ever touch that entry again;
1543          * we'll eliminate all paths to it before dropping sysctl_lock
1544          */
1545         if (unlikely(p->used)) {
1546                 struct completion wait;
1547                 init_completion(&wait);
1548                 p->unregistering = &wait;
1549                 spin_unlock(&sysctl_lock);
1550                 wait_for_completion(&wait);
1551                 spin_lock(&sysctl_lock);
1552         } else {
1553                 /* anything non-NULL; we'll never dereference it */
1554                 p->unregistering = ERR_PTR(-EINVAL);
1555         }
1556         /*
1557          * do not remove from the list until nobody holds it; walking the
1558          * list in do_sysctl() relies on that.
1559          */
1560         list_del_init(&p->ctl_entry);
1561 }
1562
1563 void sysctl_head_get(struct ctl_table_header *head)
1564 {
1565         spin_lock(&sysctl_lock);
1566         head->count++;
1567         spin_unlock(&sysctl_lock);
1568 }
1569
1570 void sysctl_head_put(struct ctl_table_header *head)
1571 {
1572         spin_lock(&sysctl_lock);
1573         if (!--head->count)
1574                 kfree(head);
1575         spin_unlock(&sysctl_lock);
1576 }
1577
1578 struct ctl_table_header *sysctl_head_grab(struct ctl_table_header *head)
1579 {
1580         if (!head)
1581                 BUG();
1582         spin_lock(&sysctl_lock);
1583         if (!use_table(head))
1584                 head = ERR_PTR(-ENOENT);
1585         spin_unlock(&sysctl_lock);
1586         return head;
1587 }
1588
1589 void sysctl_head_finish(struct ctl_table_header *head)
1590 {
1591         if (!head)
1592                 return;
1593         spin_lock(&sysctl_lock);
1594         unuse_table(head);
1595         spin_unlock(&sysctl_lock);
1596 }
1597
1598 static struct ctl_table_set *
1599 lookup_header_set(struct ctl_table_root *root, struct nsproxy *namespaces)
1600 {
1601         struct ctl_table_set *set = &root->default_set;
1602         if (root->lookup)
1603                 set = root->lookup(root, namespaces);
1604         return set;
1605 }
1606
1607 static struct list_head *
1608 lookup_header_list(struct ctl_table_root *root, struct nsproxy *namespaces)
1609 {
1610         struct ctl_table_set *set = lookup_header_set(root, namespaces);
1611         return &set->list;
1612 }
1613
1614 struct ctl_table_header *__sysctl_head_next(struct nsproxy *namespaces,
1615                                             struct ctl_table_header *prev)
1616 {
1617         struct ctl_table_root *root;
1618         struct list_head *header_list;
1619         struct ctl_table_header *head;
1620         struct list_head *tmp;
1621
1622         spin_lock(&sysctl_lock);
1623         if (prev) {
1624                 head = prev;
1625                 tmp = &prev->ctl_entry;
1626                 unuse_table(prev);
1627                 goto next;
1628         }
1629         tmp = &root_table_header.ctl_entry;
1630         for (;;) {
1631                 head = list_entry(tmp, struct ctl_table_header, ctl_entry);
1632
1633                 if (!use_table(head))
1634                         goto next;
1635                 spin_unlock(&sysctl_lock);
1636                 return head;
1637         next:
1638                 root = head->root;
1639                 tmp = tmp->next;
1640                 header_list = lookup_header_list(root, namespaces);
1641                 if (tmp != header_list)
1642                         continue;
1643
1644                 do {
1645                         root = list_entry(root->root_list.next,
1646                                         struct ctl_table_root, root_list);
1647                         if (root == &sysctl_table_root)
1648                                 goto out;
1649                         header_list = lookup_header_list(root, namespaces);
1650                 } while (list_empty(header_list));
1651                 tmp = header_list->next;
1652         }
1653 out:
1654         spin_unlock(&sysctl_lock);
1655         return NULL;
1656 }
1657
1658 struct ctl_table_header *sysctl_head_next(struct ctl_table_header *prev)
1659 {
1660         return __sysctl_head_next(current->nsproxy, prev);
1661 }
1662
1663 void register_sysctl_root(struct ctl_table_root *root)
1664 {
1665         spin_lock(&sysctl_lock);
1666         list_add_tail(&root->root_list, &sysctl_table_root.root_list);
1667         spin_unlock(&sysctl_lock);
1668 }
1669
1670 /*
1671  * sysctl_perm does NOT grant the superuser all rights automatically, because
1672  * some sysctl variables are readonly even to root.
1673  */
1674
1675 static int test_perm(int mode, int op)
1676 {
1677         if (!current_euid())
1678                 mode >>= 6;
1679         else if (in_egroup_p(0))
1680                 mode >>= 3;
1681         if ((op & ~mode & (MAY_READ|MAY_WRITE|MAY_EXEC)) == 0)
1682                 return 0;
1683         return -EACCES;
1684 }
1685
1686 int sysctl_perm(struct ctl_table_root *root, struct ctl_table *table, int op)
1687 {
1688         int error;
1689         int mode;
1690
1691         error = security_sysctl(table, op & (MAY_READ | MAY_WRITE | MAY_EXEC));
1692         if (error)
1693                 return error;
1694
1695         if (root->permissions)
1696                 mode = root->permissions(root, current->nsproxy, table);
1697         else
1698                 mode = table->mode;
1699
1700         return test_perm(mode, op);
1701 }
1702
1703 static void sysctl_set_parent(struct ctl_table *parent, struct ctl_table *table)
1704 {
1705         for (; table->procname; table++) {
1706                 table->parent = parent;
1707                 if (table->child)
1708                         sysctl_set_parent(table, table->child);
1709         }
1710 }
1711
1712 static __init int sysctl_init(void)
1713 {
1714         sysctl_set_parent(NULL, root_table);
1715 #ifdef CONFIG_SYSCTL_SYSCALL_CHECK
1716         sysctl_check_table(current->nsproxy, root_table);
1717 #endif
1718         return 0;
1719 }
1720
1721 core_initcall(sysctl_init);
1722
1723 static struct ctl_table *is_branch_in(struct ctl_table *branch,
1724                                       struct ctl_table *table)
1725 {
1726         struct ctl_table *p;
1727         const char *s = branch->procname;
1728
1729         /* branch should have named subdirectory as its first element */
1730         if (!s || !branch->child)
1731                 return NULL;
1732
1733         /* ... and nothing else */
1734         if (branch[1].procname)
1735                 return NULL;
1736
1737         /* table should contain subdirectory with the same name */
1738         for (p = table; p->procname; p++) {
1739                 if (!p->child)
1740                         continue;
1741                 if (p->procname && strcmp(p->procname, s) == 0)
1742                         return p;
1743         }
1744         return NULL;
1745 }
1746
1747 /* see if attaching q to p would be an improvement */
1748 static void try_attach(struct ctl_table_header *p, struct ctl_table_header *q)
1749 {
1750         struct ctl_table *to = p->ctl_table, *by = q->ctl_table;
1751         struct ctl_table *next;
1752         int is_better = 0;
1753         int not_in_parent = !p->attached_by;
1754
1755         while ((next = is_branch_in(by, to)) != NULL) {
1756                 if (by == q->attached_by)
1757                         is_better = 1;
1758                 if (to == p->attached_by)
1759                         not_in_parent = 1;
1760                 by = by->child;
1761                 to = next->child;
1762         }
1763
1764         if (is_better && not_in_parent) {
1765                 q->attached_by = by;
1766                 q->attached_to = to;
1767                 q->parent = p;
1768         }
1769 }
1770
1771 /**
1772  * __register_sysctl_paths - register a sysctl hierarchy
1773  * @root: List of sysctl headers to register on
1774  * @namespaces: Data to compute which lists of sysctl entries are visible
1775  * @path: The path to the directory the sysctl table is in.
1776  * @table: the top-level table structure
1777  *
1778  * Register a sysctl table hierarchy. @table should be a filled in ctl_table
1779  * array. A completely 0 filled entry terminates the table.
1780  *
1781  * The members of the &struct ctl_table structure are used as follows:
1782  *
1783  * procname - the name of the sysctl file under /proc/sys. Set to %NULL to not
1784  *            enter a sysctl file
1785  *
1786  * data - a pointer to data for use by proc_handler
1787  *
1788  * maxlen - the maximum size in bytes of the data
1789  *
1790  * mode - the file permissions for the /proc/sys file, and for sysctl(2)
1791  *
1792  * child - a pointer to the child sysctl table if this entry is a directory, or
1793  *         %NULL.
1794  *
1795  * proc_handler - the text handler routine (described below)
1796  *
1797  * de - for internal use by the sysctl routines
1798  *
1799  * extra1, extra2 - extra pointers usable by the proc handler routines
1800  *
1801  * Leaf nodes in the sysctl tree will be represented by a single file
1802  * under /proc; non-leaf nodes will be represented by directories.
1803  *
1804  * sysctl(2) can automatically manage read and write requests through
1805  * the sysctl table.  The data and maxlen fields of the ctl_table
1806  * struct enable minimal validation of the values being written to be
1807  * performed, and the mode field allows minimal authentication.
1808  *
1809  * There must be a proc_handler routine for any terminal nodes
1810  * mirrored under /proc/sys (non-terminals are handled by a built-in
1811  * directory handler).  Several default handlers are available to
1812  * cover common cases -
1813  *
1814  * proc_dostring(), proc_dointvec(), proc_dointvec_jiffies(),
1815  * proc_dointvec_userhz_jiffies(), proc_dointvec_minmax(),
1816  * proc_doulongvec_ms_jiffies_minmax(), proc_doulongvec_minmax()
1817  *
1818  * It is the handler's job to read the input buffer from user memory
1819  * and process it. The handler should return 0 on success.
1820  *
1821  * This routine returns %NULL on a failure to register, and a pointer
1822  * to the table header on success.
1823  */
1824 struct ctl_table_header *__register_sysctl_paths(
1825         struct ctl_table_root *root,
1826         struct nsproxy *namespaces,
1827         const struct ctl_path *path, struct ctl_table *table)
1828 {
1829         struct ctl_table_header *header;
1830         struct ctl_table *new, **prevp;
1831         unsigned int n, npath;
1832         struct ctl_table_set *set;
1833
1834         /* Count the path components */
1835         for (npath = 0; path[npath].procname; ++npath)
1836                 ;
1837
1838         /*
1839          * For each path component, allocate a 2-element ctl_table array.
1840          * The first array element will be filled with the sysctl entry
1841          * for this, the second will be the sentinel (procname == 0).
1842          *
1843          * We allocate everything in one go so that we don't have to
1844          * worry about freeing additional memory in unregister_sysctl_table.
1845          */
1846         header = kzalloc(sizeof(struct ctl_table_header) +
1847                          (2 * npath * sizeof(struct ctl_table)), GFP_KERNEL);
1848         if (!header)
1849                 return NULL;
1850
1851         new = (struct ctl_table *) (header + 1);
1852
1853         /* Now connect the dots */
1854         prevp = &header->ctl_table;
1855         for (n = 0; n < npath; ++n, ++path) {
1856                 /* Copy the procname */
1857                 new->procname = path->procname;
1858                 new->mode     = 0555;
1859
1860                 *prevp = new;
1861                 prevp = &new->child;
1862
1863                 new += 2;
1864         }
1865         *prevp = table;
1866         header->ctl_table_arg = table;
1867
1868         INIT_LIST_HEAD(&header->ctl_entry);
1869         header->used = 0;
1870         header->unregistering = NULL;
1871         header->root = root;
1872         sysctl_set_parent(NULL, header->ctl_table);
1873         header->count = 1;
1874 #ifdef CONFIG_SYSCTL_SYSCALL_CHECK
1875         if (sysctl_check_table(namespaces, header->ctl_table)) {
1876                 kfree(header);
1877                 return NULL;
1878         }
1879 #endif
1880         spin_lock(&sysctl_lock);
1881         header->set = lookup_header_set(root, namespaces);
1882         header->attached_by = header->ctl_table;
1883         header->attached_to = root_table;
1884         header->parent = &root_table_header;
1885         for (set = header->set; set; set = set->parent) {
1886                 struct ctl_table_header *p;
1887                 list_for_each_entry(p, &set->list, ctl_entry) {
1888                         if (p->unregistering)
1889                                 continue;
1890                         try_attach(p, header);
1891                 }
1892         }
1893         header->parent->count++;
1894         list_add_tail(&header->ctl_entry, &header->set->list);
1895         spin_unlock(&sysctl_lock);
1896
1897         return header;
1898 }
1899
1900 /**
1901  * register_sysctl_table_path - register a sysctl table hierarchy
1902  * @path: The path to the directory the sysctl table is in.
1903  * @table: the top-level table structure
1904  *
1905  * Register a sysctl table hierarchy. @table should be a filled in ctl_table
1906  * array. A completely 0 filled entry terminates the table.
1907  *
1908  * See __register_sysctl_paths for more details.
1909  */
1910 struct ctl_table_header *register_sysctl_paths(const struct ctl_path *path,
1911                                                 struct ctl_table *table)
1912 {
1913         return __register_sysctl_paths(&sysctl_table_root, current->nsproxy,
1914                                         path, table);
1915 }
1916
1917 /**
1918  * register_sysctl_table - register a sysctl table hierarchy
1919  * @table: the top-level table structure
1920  *
1921  * Register a sysctl table hierarchy. @table should be a filled in ctl_table
1922  * array. A completely 0 filled entry terminates the table.
1923  *
1924  * See register_sysctl_paths for more details.
1925  */
1926 struct ctl_table_header *register_sysctl_table(struct ctl_table *table)
1927 {
1928         static const struct ctl_path null_path[] = { {} };
1929
1930         return register_sysctl_paths(null_path, table);
1931 }
1932
1933 /**
1934  * unregister_sysctl_table - unregister a sysctl table hierarchy
1935  * @header: the header returned from register_sysctl_table
1936  *
1937  * Unregisters the sysctl table and all children. proc entries may not
1938  * actually be removed until they are no longer used by anyone.
1939  */
1940 void unregister_sysctl_table(struct ctl_table_header * header)
1941 {
1942         might_sleep();
1943
1944         if (header == NULL)
1945                 return;
1946
1947         spin_lock(&sysctl_lock);
1948         start_unregistering(header);
1949         if (!--header->parent->count) {
1950                 WARN_ON(1);
1951                 kfree(header->parent);
1952         }
1953         if (!--header->count)
1954                 kfree(header);
1955         spin_unlock(&sysctl_lock);
1956 }
1957
1958 int sysctl_is_seen(struct ctl_table_header *p)
1959 {
1960         struct ctl_table_set *set = p->set;
1961         int res;
1962         spin_lock(&sysctl_lock);
1963         if (p->unregistering)
1964                 res = 0;
1965         else if (!set->is_seen)
1966                 res = 1;
1967         else
1968                 res = set->is_seen(set);
1969         spin_unlock(&sysctl_lock);
1970         return res;
1971 }
1972
1973 void setup_sysctl_set(struct ctl_table_set *p,
1974         struct ctl_table_set *parent,
1975         int (*is_seen)(struct ctl_table_set *))
1976 {
1977         INIT_LIST_HEAD(&p->list);
1978         p->parent = parent ? parent : &sysctl_table_root.default_set;
1979         p->is_seen = is_seen;
1980 }
1981
1982 #else /* !CONFIG_SYSCTL */
1983 struct ctl_table_header *register_sysctl_table(struct ctl_table * table)
1984 {
1985         return NULL;
1986 }
1987
1988 struct ctl_table_header *register_sysctl_paths(const struct ctl_path *path,
1989                                                     struct ctl_table *table)
1990 {
1991         return NULL;
1992 }
1993
1994 void unregister_sysctl_table(struct ctl_table_header * table)
1995 {
1996 }
1997
1998 void setup_sysctl_set(struct ctl_table_set *p,
1999         struct ctl_table_set *parent,
2000         int (*is_seen)(struct ctl_table_set *))
2001 {
2002 }
2003
2004 void sysctl_head_put(struct ctl_table_header *head)
2005 {
2006 }
2007
2008 #endif /* CONFIG_SYSCTL */
2009
2010 /*
2011  * /proc/sys support
2012  */
2013
2014 #ifdef CONFIG_PROC_SYSCTL
2015
2016 static int _proc_do_string(void* data, int maxlen, int write,
2017                            void __user *buffer,
2018                            size_t *lenp, loff_t *ppos)
2019 {
2020         size_t len;
2021         char __user *p;
2022         char c;
2023
2024         if (!data || !maxlen || !*lenp) {
2025                 *lenp = 0;
2026                 return 0;
2027         }
2028
2029         if (write) {
2030                 len = 0;
2031                 p = buffer;
2032                 while (len < *lenp) {
2033                         if (get_user(c, p++))
2034                                 return -EFAULT;
2035                         if (c == 0 || c == '\n')
2036                                 break;
2037                         len++;
2038                 }
2039                 if (len >= maxlen)
2040                         len = maxlen-1;
2041                 if(copy_from_user(data, buffer, len))
2042                         return -EFAULT;
2043                 ((char *) data)[len] = 0;
2044                 *ppos += *lenp;
2045         } else {
2046                 len = strlen(data);
2047                 if (len > maxlen)
2048                         len = maxlen;
2049
2050                 if (*ppos > len) {
2051                         *lenp = 0;
2052                         return 0;
2053                 }
2054
2055                 data += *ppos;
2056                 len  -= *ppos;
2057
2058                 if (len > *lenp)
2059                         len = *lenp;
2060                 if (len)
2061                         if(copy_to_user(buffer, data, len))
2062                                 return -EFAULT;
2063                 if (len < *lenp) {
2064                         if(put_user('\n', ((char __user *) buffer) + len))
2065                                 return -EFAULT;
2066                         len++;
2067                 }
2068                 *lenp = len;
2069                 *ppos += len;
2070         }
2071         return 0;
2072 }
2073
2074 /**
2075  * proc_dostring - read a string sysctl
2076  * @table: the sysctl table
2077  * @write: %TRUE if this is a write to the sysctl file
2078  * @buffer: the user buffer
2079  * @lenp: the size of the user buffer
2080  * @ppos: file position
2081  *
2082  * Reads/writes a string from/to the user buffer. If the kernel
2083  * buffer provided is not large enough to hold the string, the
2084  * string is truncated. The copied string is %NULL-terminated.
2085  * If the string is being read by the user process, it is copied
2086  * and a newline '\n' is added. It is truncated if the buffer is
2087  * not large enough.
2088  *
2089  * Returns 0 on success.
2090  */
2091 int proc_dostring(struct ctl_table *table, int write,
2092                   void __user *buffer, size_t *lenp, loff_t *ppos)
2093 {
2094         return _proc_do_string(table->data, table->maxlen, write,
2095                                buffer, lenp, ppos);
2096 }
2097
2098 static size_t proc_skip_spaces(char **buf)
2099 {
2100         size_t ret;
2101         char *tmp = skip_spaces(*buf);
2102         ret = tmp - *buf;
2103         *buf = tmp;
2104         return ret;
2105 }
2106
2107 static void proc_skip_char(char **buf, size_t *size, const char v)
2108 {
2109         while (*size) {
2110                 if (**buf != v)
2111                         break;
2112                 (*size)--;
2113                 (*buf)++;
2114         }
2115 }
2116
2117 #define TMPBUFLEN 22
2118 /**
2119  * proc_get_long - reads an ASCII formatted integer from a user buffer
2120  *
2121  * @buf: a kernel buffer
2122  * @size: size of the kernel buffer
2123  * @val: this is where the number will be stored
2124  * @neg: set to %TRUE if number is negative
2125  * @perm_tr: a vector which contains the allowed trailers
2126  * @perm_tr_len: size of the perm_tr vector
2127  * @tr: pointer to store the trailer character
2128  *
2129  * In case of success %0 is returned and @buf and @size are updated with
2130  * the amount of bytes read. If @tr is non-NULL and a trailing
2131  * character exists (size is non-zero after returning from this
2132  * function), @tr is updated with the trailing character.
2133  */
2134 static int proc_get_long(char **buf, size_t *size,
2135                           unsigned long *val, bool *neg,
2136                           const char *perm_tr, unsigned perm_tr_len, char *tr)
2137 {
2138         int len;
2139         char *p, tmp[TMPBUFLEN];
2140
2141         if (!*size)
2142                 return -EINVAL;
2143
2144         len = *size;
2145         if (len > TMPBUFLEN - 1)
2146                 len = TMPBUFLEN - 1;
2147
2148         memcpy(tmp, *buf, len);
2149
2150         tmp[len] = 0;
2151         p = tmp;
2152         if (*p == '-' && *size > 1) {
2153                 *neg = true;
2154                 p++;
2155         } else
2156                 *neg = false;
2157         if (!isdigit(*p))
2158                 return -EINVAL;
2159
2160         *val = simple_strtoul(p, &p, 0);
2161
2162         len = p - tmp;
2163
2164         /* We don't know if the next char is whitespace thus we may accept
2165          * invalid integers (e.g. 1234...a) or two integers instead of one
2166          * (e.g. 123...1). So lets not allow such large numbers. */
2167         if (len == TMPBUFLEN - 1)
2168                 return -EINVAL;
2169
2170         if (len < *size && perm_tr_len && !memchr(perm_tr, *p, perm_tr_len))
2171                 return -EINVAL;
2172
2173         if (tr && (len < *size))
2174                 *tr = *p;
2175
2176         *buf += len;
2177         *size -= len;
2178
2179         return 0;
2180 }
2181
2182 /**
2183  * proc_put_long - converts an integer to a decimal ASCII formatted string
2184  *
2185  * @buf: the user buffer
2186  * @size: the size of the user buffer
2187  * @val: the integer to be converted
2188  * @neg: sign of the number, %TRUE for negative
2189  *
2190  * In case of success %0 is returned and @buf and @size are updated with
2191  * the amount of bytes written.
2192  */
2193 static int proc_put_long(void __user **buf, size_t *size, unsigned long val,
2194                           bool neg)
2195 {
2196         int len;
2197         char tmp[TMPBUFLEN], *p = tmp;
2198
2199         sprintf(p, "%s%lu", neg ? "-" : "", val);
2200         len = strlen(tmp);
2201         if (len > *size)
2202                 len = *size;
2203         if (copy_to_user(*buf, tmp, len))
2204                 return -EFAULT;
2205         *size -= len;
2206         *buf += len;
2207         return 0;
2208 }
2209 #undef TMPBUFLEN
2210
2211 static int proc_put_char(void __user **buf, size_t *size, char c)
2212 {
2213         if (*size) {
2214                 char __user **buffer = (char __user **)buf;
2215                 if (put_user(c, *buffer))
2216                         return -EFAULT;
2217                 (*size)--, (*buffer)++;
2218                 *buf = *buffer;
2219         }
2220         return 0;
2221 }
2222
2223 static int do_proc_dointvec_conv(bool *negp, unsigned long *lvalp,
2224                                  int *valp,
2225                                  int write, void *data)
2226 {
2227         if (write) {
2228                 *valp = *negp ? -*lvalp : *lvalp;
2229         } else {
2230                 int val = *valp;
2231                 if (val < 0) {
2232                         *negp = true;
2233                         *lvalp = (unsigned long)-val;
2234                 } else {
2235                         *negp = false;
2236                         *lvalp = (unsigned long)val;
2237                 }
2238         }
2239         return 0;
2240 }
2241
2242 static const char proc_wspace_sep[] = { ' ', '\t', '\n' };
2243
2244 static int __do_proc_dointvec(void *tbl_data, struct ctl_table *table,
2245                   int write, void __user *buffer,
2246                   size_t *lenp, loff_t *ppos,
2247                   int (*conv)(bool *negp, unsigned long *lvalp, int *valp,
2248                               int write, void *data),
2249                   void *data)
2250 {
2251         int *i, vleft, first = 1, err = 0;
2252         unsigned long page = 0;
2253         size_t left;
2254         char *kbuf;
2255
2256         if (!tbl_data || !table->maxlen || !*lenp || (*ppos && !write)) {
2257                 *lenp = 0;
2258                 return 0;
2259         }
2260
2261         i = (int *) tbl_data;
2262         vleft = table->maxlen / sizeof(*i);
2263         left = *lenp;
2264
2265         if (!conv)
2266                 conv = do_proc_dointvec_conv;
2267
2268         if (write) {
2269                 if (left > PAGE_SIZE - 1)
2270                         left = PAGE_SIZE - 1;
2271                 page = __get_free_page(GFP_TEMPORARY);
2272                 kbuf = (char *) page;
2273                 if (!kbuf)
2274                         return -ENOMEM;
2275                 if (copy_from_user(kbuf, buffer, left)) {
2276                         err = -EFAULT;
2277                         goto free;
2278                 }
2279                 kbuf[left] = 0;
2280         }
2281
2282         for (; left && vleft--; i++, first=0) {
2283                 unsigned long lval;
2284                 bool neg;
2285
2286                 if (write) {
2287                         left -= proc_skip_spaces(&kbuf);
2288
2289                         if (!left)
2290                                 break;
2291                         err = proc_get_long(&kbuf, &left, &lval, &neg,
2292                                              proc_wspace_sep,
2293                                              sizeof(proc_wspace_sep), NULL);
2294                         if (err)
2295                                 break;
2296                         if (conv(&neg, &lval, i, 1, data)) {
2297                                 err = -EINVAL;
2298                                 break;
2299                         }
2300                 } else {
2301                         if (conv(&neg, &lval, i, 0, data)) {
2302                                 err = -EINVAL;
2303                                 break;
2304                         }
2305                         if (!first)
2306                                 err = proc_put_char(&buffer, &left, '\t');
2307                         if (err)
2308                                 break;
2309                         err = proc_put_long(&buffer, &left, lval, neg);
2310                         if (err)
2311                                 break;
2312                 }
2313         }
2314
2315         if (!write && !first && left && !err)
2316                 err = proc_put_char(&buffer, &left, '\n');
2317         if (write && !err && left)
2318                 left -= proc_skip_spaces(&kbuf);
2319 free:
2320         if (write) {
2321                 free_page(page);
2322                 if (first)
2323                         return err ? : -EINVAL;
2324         }
2325         *lenp -= left;
2326         *ppos += *lenp;
2327         return err;
2328 }
2329
2330 static int do_proc_dointvec(struct ctl_table *table, int write,
2331                   void __user *buffer, size_t *lenp, loff_t *ppos,
2332                   int (*conv)(bool *negp, unsigned long *lvalp, int *valp,
2333                               int write, void *data),
2334                   void *data)
2335 {
2336         return __do_proc_dointvec(table->data, table, write,
2337                         buffer, lenp, ppos, conv, data);
2338 }
2339
2340 /**
2341  * proc_dointvec - read a vector of integers
2342  * @table: the sysctl table
2343  * @write: %TRUE if this is a write to the sysctl file
2344  * @buffer: the user buffer
2345  * @lenp: the size of the user buffer
2346  * @ppos: file position
2347  *
2348  * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2349  * values from/to the user buffer, treated as an ASCII string.
2350  *
2351  * Returns 0 on success.
2352  */
2353 int proc_dointvec(struct ctl_table *table, int write,
2354                      void __user *buffer, size_t *lenp, loff_t *ppos)
2355 {
2356     return do_proc_dointvec(table,write,buffer,lenp,ppos,
2357                             NULL,NULL);
2358 }
2359
2360 /*
2361  * Taint values can only be increased
2362  * This means we can safely use a temporary.
2363  */
2364 static int proc_taint(struct ctl_table *table, int write,
2365                                void __user *buffer, size_t *lenp, loff_t *ppos)
2366 {
2367         struct ctl_table t;
2368         unsigned long tmptaint = get_taint();
2369         int err;
2370
2371         if (write && !capable(CAP_SYS_ADMIN))
2372                 return -EPERM;
2373
2374         t = *table;
2375         t.data = &tmptaint;
2376         err = proc_doulongvec_minmax(&t, write, buffer, lenp, ppos);
2377         if (err < 0)
2378                 return err;
2379
2380         if (write) {
2381                 /*
2382                  * Poor man's atomic or. Not worth adding a primitive
2383                  * to everyone's atomic.h for this
2384                  */
2385                 int i;
2386                 for (i = 0; i < BITS_PER_LONG && tmptaint >> i; i++) {
2387                         if ((tmptaint >> i) & 1)
2388                                 add_taint(i);
2389                 }
2390         }
2391
2392         return err;
2393 }
2394
2395 struct do_proc_dointvec_minmax_conv_param {
2396         int *min;
2397         int *max;
2398 };
2399
2400 static int do_proc_dointvec_minmax_conv(bool *negp, unsigned long *lvalp,
2401                                         int *valp,
2402                                         int write, void *data)
2403 {
2404         struct do_proc_dointvec_minmax_conv_param *param = data;
2405         if (write) {
2406                 int val = *negp ? -*lvalp : *lvalp;
2407                 if ((param->min && *param->min > val) ||
2408                     (param->max && *param->max < val))
2409                         return -EINVAL;
2410                 *valp = val;
2411         } else {
2412                 int val = *valp;
2413                 if (val < 0) {
2414                         *negp = true;
2415                         *lvalp = (unsigned long)-val;
2416                 } else {
2417                         *negp = false;
2418                         *lvalp = (unsigned long)val;
2419                 }
2420         }
2421         return 0;
2422 }
2423
2424 /**
2425  * proc_dointvec_minmax - read a vector of integers with min/max values
2426  * @table: the sysctl table
2427  * @write: %TRUE if this is a write to the sysctl file
2428  * @buffer: the user buffer
2429  * @lenp: the size of the user buffer
2430  * @ppos: file position
2431  *
2432  * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2433  * values from/to the user buffer, treated as an ASCII string.
2434  *
2435  * This routine will ensure the values are within the range specified by
2436  * table->extra1 (min) and table->extra2 (max).
2437  *
2438  * Returns 0 on success.
2439  */
2440 int proc_dointvec_minmax(struct ctl_table *table, int write,
2441                   void __user *buffer, size_t *lenp, loff_t *ppos)
2442 {
2443         struct do_proc_dointvec_minmax_conv_param param = {
2444                 .min = (int *) table->extra1,
2445                 .max = (int *) table->extra2,
2446         };
2447         return do_proc_dointvec(table, write, buffer, lenp, ppos,
2448                                 do_proc_dointvec_minmax_conv, &param);
2449 }
2450
2451 static int __do_proc_doulongvec_minmax(void *data, struct ctl_table *table, int write,
2452                                      void __user *buffer,
2453                                      size_t *lenp, loff_t *ppos,
2454                                      unsigned long convmul,
2455                                      unsigned long convdiv)
2456 {
2457         unsigned long *i, *min, *max;
2458         int vleft, first = 1, err = 0;
2459         unsigned long page = 0;
2460         size_t left;
2461         char *kbuf;
2462
2463         if (!data || !table->maxlen || !*lenp || (*ppos && !write)) {
2464                 *lenp = 0;
2465                 return 0;
2466         }
2467
2468         i = (unsigned long *) data;
2469         min = (unsigned long *) table->extra1;
2470         max = (unsigned long *) table->extra2;
2471         vleft = table->maxlen / sizeof(unsigned long);
2472         left = *lenp;
2473
2474         if (write) {
2475                 if (left > PAGE_SIZE - 1)
2476                         left = PAGE_SIZE - 1;
2477                 page = __get_free_page(GFP_TEMPORARY);
2478                 kbuf = (char *) page;
2479                 if (!kbuf)
2480                         return -ENOMEM;
2481                 if (copy_from_user(kbuf, buffer, left)) {
2482                         err = -EFAULT;
2483                         goto free;
2484                 }
2485                 kbuf[left] = 0;
2486         }
2487
2488         for (; left && vleft--; i++, first = 0) {
2489                 unsigned long val;
2490
2491                 if (write) {
2492                         bool neg;
2493
2494                         left -= proc_skip_spaces(&kbuf);
2495
2496                         err = proc_get_long(&kbuf, &left, &val, &neg,
2497                                              proc_wspace_sep,
2498                                              sizeof(proc_wspace_sep), NULL);
2499                         if (err)
2500                                 break;
2501                         if (neg)
2502                                 continue;
2503                         if ((min && val < *min) || (max && val > *max))
2504                                 continue;
2505                         *i = val;
2506                 } else {
2507                         val = convdiv * (*i) / convmul;
2508                         if (!first)
2509                                 err = proc_put_char(&buffer, &left, '\t');
2510                         err = proc_put_long(&buffer, &left, val, false);
2511                         if (err)
2512                                 break;
2513                 }
2514         }
2515
2516         if (!write && !first && left && !err)
2517                 err = proc_put_char(&buffer, &left, '\n');
2518         if (write && !err)
2519                 left -= proc_skip_spaces(&kbuf);
2520 free:
2521         if (write) {
2522                 free_page(page);
2523                 if (first)
2524                         return err ? : -EINVAL;
2525         }
2526         *lenp -= left;
2527         *ppos += *lenp;
2528         return err;
2529 }
2530
2531 static int do_proc_doulongvec_minmax(struct ctl_table *table, int write,
2532                                      void __user *buffer,
2533                                      size_t *lenp, loff_t *ppos,
2534                                      unsigned long convmul,
2535                                      unsigned long convdiv)
2536 {
2537         return __do_proc_doulongvec_minmax(table->data, table, write,
2538                         buffer, lenp, ppos, convmul, convdiv);
2539 }
2540
2541 /**
2542  * proc_doulongvec_minmax - read a vector of long integers with min/max values
2543  * @table: the sysctl table
2544  * @write: %TRUE if this is a write to the sysctl file
2545  * @buffer: the user buffer
2546  * @lenp: the size of the user buffer
2547  * @ppos: file position
2548  *
2549  * Reads/writes up to table->maxlen/sizeof(unsigned long) unsigned long
2550  * values from/to the user buffer, treated as an ASCII string.
2551  *
2552  * This routine will ensure the values are within the range specified by
2553  * table->extra1 (min) and table->extra2 (max).
2554  *
2555  * Returns 0 on success.
2556  */
2557 int proc_doulongvec_minmax(struct ctl_table *table, int write,
2558                            void __user *buffer, size_t *lenp, loff_t *ppos)
2559 {
2560     return do_proc_doulongvec_minmax(table, write, buffer, lenp, ppos, 1l, 1l);
2561 }
2562
2563 /**
2564  * proc_doulongvec_ms_jiffies_minmax - read a vector of millisecond values 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. The values
2573  * are treated as milliseconds, and converted to jiffies when they are stored.
2574  *
2575  * This routine will ensure the values are within the range specified by
2576  * table->extra1 (min) and table->extra2 (max).
2577  *
2578  * Returns 0 on success.
2579  */
2580 int proc_doulongvec_ms_jiffies_minmax(struct ctl_table *table, int write,
2581                                       void __user *buffer,
2582                                       size_t *lenp, loff_t *ppos)
2583 {
2584     return do_proc_doulongvec_minmax(table, write, buffer,
2585                                      lenp, ppos, HZ, 1000l);
2586 }
2587
2588
2589 static int do_proc_dointvec_jiffies_conv(bool *negp, unsigned long *lvalp,
2590                                          int *valp,
2591                                          int write, void *data)
2592 {
2593         if (write) {
2594                 if (*lvalp > LONG_MAX / HZ)
2595                         return 1;
2596                 *valp = *negp ? -(*lvalp*HZ) : (*lvalp*HZ);
2597         } else {
2598                 int val = *valp;
2599                 unsigned long lval;
2600                 if (val < 0) {
2601                         *negp = true;
2602                         lval = (unsigned long)-val;
2603                 } else {
2604                         *negp = false;
2605                         lval = (unsigned long)val;
2606                 }
2607                 *lvalp = lval / HZ;
2608         }
2609         return 0;
2610 }
2611
2612 static int do_proc_dointvec_userhz_jiffies_conv(bool *negp, unsigned long *lvalp,
2613                                                 int *valp,
2614                                                 int write, void *data)
2615 {
2616         if (write) {
2617                 if (USER_HZ < HZ && *lvalp > (LONG_MAX / HZ) * USER_HZ)
2618                         return 1;
2619                 *valp = clock_t_to_jiffies(*negp ? -*lvalp : *lvalp);
2620         } else {
2621                 int val = *valp;
2622                 unsigned long lval;
2623                 if (val < 0) {
2624                         *negp = true;
2625                         lval = (unsigned long)-val;
2626                 } else {
2627                         *negp = false;
2628                         lval = (unsigned long)val;
2629                 }
2630                 *lvalp = jiffies_to_clock_t(lval);
2631         }
2632         return 0;
2633 }
2634
2635 static int do_proc_dointvec_ms_jiffies_conv(bool *negp, unsigned long *lvalp,
2636                                             int *valp,
2637                                             int write, void *data)
2638 {
2639         if (write) {
2640                 *valp = msecs_to_jiffies(*negp ? -*lvalp : *lvalp);
2641         } else {
2642                 int val = *valp;
2643                 unsigned long lval;
2644                 if (val < 0) {
2645                         *negp = true;
2646                         lval = (unsigned long)-val;
2647                 } else {
2648                         *negp = false;
2649                         lval = (unsigned long)val;
2650                 }
2651                 *lvalp = jiffies_to_msecs(lval);
2652         }
2653         return 0;
2654 }
2655
2656 /**
2657  * proc_dointvec_jiffies - read a vector of integers as seconds
2658  * @table: the sysctl table
2659  * @write: %TRUE if this is a write to the sysctl file
2660  * @buffer: the user buffer
2661  * @lenp: the size of the user buffer
2662  * @ppos: file position
2663  *
2664  * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2665  * values from/to the user buffer, treated as an ASCII string.
2666  * The values read are assumed to be in seconds, and are converted into
2667  * jiffies.
2668  *
2669  * Returns 0 on success.
2670  */
2671 int proc_dointvec_jiffies(struct ctl_table *table, int write,
2672                           void __user *buffer, size_t *lenp, loff_t *ppos)
2673 {
2674     return do_proc_dointvec(table,write,buffer,lenp,ppos,
2675                             do_proc_dointvec_jiffies_conv,NULL);
2676 }
2677
2678 /**
2679  * proc_dointvec_userhz_jiffies - read a vector of integers as 1/USER_HZ 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: pointer to the 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 1/USER_HZ seconds, and
2689  * are converted into jiffies.
2690  *
2691  * Returns 0 on success.
2692  */
2693 int proc_dointvec_userhz_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_userhz_jiffies_conv,NULL);
2698 }
2699
2700 /**
2701  * proc_dointvec_ms_jiffies - read a vector of integers as 1 milliseconds
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: file position
2707  * @ppos: the current position in the file
2708  *
2709  * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2710  * values from/to the user buffer, treated as an ASCII string.
2711  * The values read are assumed to be in 1/1000 seconds, and
2712  * are converted into jiffies.
2713  *
2714  * Returns 0 on success.
2715  */
2716 int proc_dointvec_ms_jiffies(struct ctl_table *table, int write,
2717                              void __user *buffer, size_t *lenp, loff_t *ppos)
2718 {
2719         return do_proc_dointvec(table, write, buffer, lenp, ppos,
2720                                 do_proc_dointvec_ms_jiffies_conv, NULL);
2721 }
2722
2723 static int proc_do_cad_pid(struct ctl_table *table, int write,
2724                            void __user *buffer, size_t *lenp, loff_t *ppos)
2725 {
2726         struct pid *new_pid;
2727         pid_t tmp;
2728         int r;
2729
2730         tmp = pid_vnr(cad_pid);
2731
2732         r = __do_proc_dointvec(&tmp, table, write, buffer,
2733                                lenp, ppos, NULL, NULL);
2734         if (r || !write)
2735                 return r;
2736
2737         new_pid = find_get_pid(tmp);
2738         if (!new_pid)
2739                 return -ESRCH;
2740
2741         put_pid(xchg(&cad_pid, new_pid));
2742         return 0;
2743 }
2744
2745 /**
2746  * proc_do_large_bitmap - read/write from/to a large bitmap
2747  * @table: the sysctl table
2748  * @write: %TRUE if this is a write to the sysctl file
2749  * @buffer: the user buffer
2750  * @lenp: the size of the user buffer
2751  * @ppos: file position
2752  *
2753  * The bitmap is stored at table->data and the bitmap length (in bits)
2754  * in table->maxlen.
2755  *
2756  * We use a range comma separated format (e.g. 1,3-4,10-10) so that
2757  * large bitmaps may be represented in a compact manner. Writing into
2758  * the file will clear the bitmap then update it with the given input.
2759  *
2760  * Returns 0 on success.
2761  */
2762 int proc_do_large_bitmap(struct ctl_table *table, int write,
2763                          void __user *buffer, size_t *lenp, loff_t *ppos)
2764 {
2765         int err = 0;
2766         bool first = 1;
2767         size_t left = *lenp;
2768         unsigned long bitmap_len = table->maxlen;
2769         unsigned long *bitmap = (unsigned long *) table->data;
2770         unsigned long *tmp_bitmap = NULL;
2771         char tr_a[] = { '-', ',', '\n' }, tr_b[] = { ',', '\n', 0 }, c;
2772
2773         if (!bitmap_len || !left || (*ppos && !write)) {
2774                 *lenp = 0;
2775                 return 0;
2776         }
2777
2778         if (write) {
2779                 unsigned long page = 0;
2780                 char *kbuf;
2781
2782                 if (left > PAGE_SIZE - 1)
2783                         left = PAGE_SIZE - 1;
2784
2785                 page = __get_free_page(GFP_TEMPORARY);
2786                 kbuf = (char *) page;
2787                 if (!kbuf)
2788                         return -ENOMEM;
2789                 if (copy_from_user(kbuf, buffer, left)) {
2790                         free_page(page);
2791                         return -EFAULT;
2792                 }
2793                 kbuf[left] = 0;
2794
2795                 tmp_bitmap = kzalloc(BITS_TO_LONGS(bitmap_len) * sizeof(unsigned long),
2796                                      GFP_KERNEL);
2797                 if (!tmp_bitmap) {
2798                         free_page(page);
2799                         return -ENOMEM;
2800                 }
2801                 proc_skip_char(&kbuf, &left, '\n');
2802                 while (!err && left) {
2803                         unsigned long val_a, val_b;
2804                         bool neg;
2805
2806                         err = proc_get_long(&kbuf, &left, &val_a, &neg, tr_a,
2807                                              sizeof(tr_a), &c);
2808                         if (err)
2809                                 break;
2810                         if (val_a >= bitmap_len || neg) {
2811                                 err = -EINVAL;
2812                                 break;
2813                         }
2814
2815                         val_b = val_a;
2816                         if (left) {
2817                                 kbuf++;
2818                                 left--;
2819                         }
2820
2821                         if (c == '-') {
2822                                 err = proc_get_long(&kbuf, &left, &val_b,
2823                                                      &neg, tr_b, sizeof(tr_b),
2824                                                      &c);
2825                                 if (err)
2826                                         break;
2827                                 if (val_b >= bitmap_len || neg ||
2828                                     val_a > val_b) {
2829                                         err = -EINVAL;
2830                                         break;
2831                                 }
2832                                 if (left) {
2833                                         kbuf++;
2834                                         left--;
2835                                 }
2836                         }
2837
2838                         while (val_a <= val_b)
2839                                 set_bit(val_a++, tmp_bitmap);
2840
2841                         first = 0;
2842                         proc_skip_char(&kbuf, &left, '\n');
2843                 }
2844                 free_page(page);
2845         } else {
2846                 unsigned long bit_a, bit_b = 0;
2847
2848                 while (left) {
2849                         bit_a = find_next_bit(bitmap, bitmap_len, bit_b);
2850                         if (bit_a >= bitmap_len)
2851                                 break;
2852                         bit_b = find_next_zero_bit(bitmap, bitmap_len,
2853                                                    bit_a + 1) - 1;
2854
2855                         if (!first) {
2856                                 err = proc_put_char(&buffer, &left, ',');
2857                                 if (err)
2858                                         break;
2859                         }
2860                         err = proc_put_long(&buffer, &left, bit_a, false);
2861                         if (err)
2862                                 break;
2863                         if (bit_a != bit_b) {
2864                                 err = proc_put_char(&buffer, &left, '-');
2865                                 if (err)
2866                                         break;
2867                                 err = proc_put_long(&buffer, &left, bit_b, false);
2868                                 if (err)
2869                                         break;
2870                         }
2871
2872                         first = 0; bit_b++;
2873                 }
2874                 if (!err)
2875                         err = proc_put_char(&buffer, &left, '\n');
2876         }
2877
2878         if (!err) {
2879                 if (write) {
2880                         if (*ppos)
2881                                 bitmap_or(bitmap, bitmap, tmp_bitmap, bitmap_len);
2882                         else
2883                                 memcpy(bitmap, tmp_bitmap,
2884                                         BITS_TO_LONGS(bitmap_len) * sizeof(unsigned long));
2885                 }
2886                 kfree(tmp_bitmap);
2887                 *lenp -= left;
2888                 *ppos += *lenp;
2889                 return 0;
2890         } else {
2891                 kfree(tmp_bitmap);
2892                 return err;
2893         }
2894 }
2895
2896 #else /* CONFIG_PROC_SYSCTL */
2897
2898 int proc_dostring(struct ctl_table *table, int write,
2899                   void __user *buffer, size_t *lenp, loff_t *ppos)
2900 {
2901         return -ENOSYS;
2902 }
2903
2904 int proc_dointvec(struct ctl_table *table, int write,
2905                   void __user *buffer, size_t *lenp, loff_t *ppos)
2906 {
2907         return -ENOSYS;
2908 }
2909
2910 int proc_dointvec_minmax(struct ctl_table *table, int write,
2911                     void __user *buffer, size_t *lenp, loff_t *ppos)
2912 {
2913         return -ENOSYS;
2914 }
2915
2916 int proc_dointvec_jiffies(struct ctl_table *table, int write,
2917                     void __user *buffer, size_t *lenp, loff_t *ppos)
2918 {
2919         return -ENOSYS;
2920 }
2921
2922 int proc_dointvec_userhz_jiffies(struct ctl_table *table, int write,
2923                     void __user *buffer, size_t *lenp, loff_t *ppos)
2924 {
2925         return -ENOSYS;
2926 }
2927
2928 int proc_dointvec_ms_jiffies(struct ctl_table *table, int write,
2929                              void __user *buffer, size_t *lenp, loff_t *ppos)
2930 {
2931         return -ENOSYS;
2932 }
2933
2934 int proc_doulongvec_minmax(struct ctl_table *table, int write,
2935                     void __user *buffer, size_t *lenp, loff_t *ppos)
2936 {
2937         return -ENOSYS;
2938 }
2939
2940 int proc_doulongvec_ms_jiffies_minmax(struct ctl_table *table, int write,
2941                                       void __user *buffer,
2942                                       size_t *lenp, loff_t *ppos)
2943 {
2944     return -ENOSYS;
2945 }
2946
2947
2948 #endif /* CONFIG_PROC_SYSCTL */
2949
2950 /*
2951  * No sense putting this after each symbol definition, twice,
2952  * exception granted :-)
2953  */
2954 EXPORT_SYMBOL(proc_dointvec);
2955 EXPORT_SYMBOL(proc_dointvec_jiffies);
2956 EXPORT_SYMBOL(proc_dointvec_minmax);
2957 EXPORT_SYMBOL(proc_dointvec_userhz_jiffies);
2958 EXPORT_SYMBOL(proc_dointvec_ms_jiffies);
2959 EXPORT_SYMBOL(proc_dostring);
2960 EXPORT_SYMBOL(proc_doulongvec_minmax);
2961 EXPORT_SYMBOL(proc_doulongvec_ms_jiffies_minmax);
2962 EXPORT_SYMBOL(register_sysctl_table);
2963 EXPORT_SYMBOL(register_sysctl_paths);
2964 EXPORT_SYMBOL(unregister_sysctl_table);