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_PRINTK
 173 static int proc_dmesg_restrict(struct ctl_table *table, int write,
 174                                 void __user *buffer, size_t *lenp, loff_t *ppos);
 175 #endif
 176
 177 #ifdef CONFIG_MAGIC_SYSRQ
 178 /* Note: sysrq code uses it's own private copy */
 179 static int __sysrq_enabled = SYSRQ_DEFAULT_ENABLE;
 180
 181 static int sysrq_sysctl_handler(ctl_table *table, int write,
 182                                 void __user *buffer, size_t *lenp,
 183                                 loff_t *ppos)
 184 {
 185         int error;
 186
 187         error = proc_dointvec(table, write, buffer, lenp, ppos);
 188         if (error)
 189                 return error;
 190
 191         if (write)
 192                 sysrq_toggle_support(__sysrq_enabled);
 193
 194         return 0;
 195 }
 196
 197 #endif
 198
 199 static struct ctl_table root_table[];
 200 static struct ctl_table_root sysctl_table_root;
 201 static struct ctl_table_header root_table_header = {
 202         {{.count = 1,
 203         .ctl_table = root_table,
 204         .ctl_entry = LIST_HEAD_INIT(sysctl_table_root.default_set.list),}},
 205         .root = &sysctl_table_root,
 206         .set = &sysctl_table_root.default_set,
 207 };
 208 static struct ctl_table_root sysctl_table_root = {
 209         .root_list = LIST_HEAD_INIT(sysctl_table_root.root_list),
 210         .default_set.list = LIST_HEAD_INIT(root_table_header.ctl_entry),
 211 };
 212
 213 static struct ctl_table kern_table[];
 214 static struct ctl_table vm_table[];
 215 static struct ctl_table fs_table[];
 216 static struct ctl_table debug_table[];
 217 static struct ctl_table dev_table[];
 218 extern struct ctl_table random_table[];
 219 #ifdef CONFIG_EPOLL
 220 extern struct ctl_table epoll_table[];
 221 #endif
 222
 223 #ifdef HAVE_ARCH_PICK_MMAP_LAYOUT
 224 int sysctl_legacy_va_layout;
 225 #endif
 226
 227 /* The default sysctl tables: */
 228
 229 static struct ctl_table root_table[] = {
 230         {
 231                 .procname       = "kernel",
 232                 .mode           = 0555,
 233                 .child          = kern_table,
 234         },
 235         {
 236                 .procname       = "vm",
 237                 .mode           = 0555,
 238                 .child          = vm_table,
 239         },
 240         {
 241                 .procname       = "fs",
 242                 .mode           = 0555,
 243                 .child          = fs_table,
 244         },
 245         {
 246                 .procname       = "debug",
 247                 .mode           = 0555,
 248                 .child          = debug_table,
 249         },
 250         {
 251                 .procname       = "dev",
 252                 .mode           = 0555,
 253                 .child          = dev_table,
 254         },
 255         { }
 256 };
 257
 258 #ifdef CONFIG_SCHED_DEBUG
 259 static int min_sched_granularity_ns = 100000;           /* 100 usecs */
 260 static int max_sched_granularity_ns = NSEC_PER_SEC;     /* 1 second */
 261 static int min_wakeup_granularity_ns;                   /* 0 usecs */
 262 static int max_wakeup_granularity_ns = NSEC_PER_SEC;    /* 1 second */
 263 static int min_sched_tunable_scaling = SCHED_TUNABLESCALING_NONE;
 264 static int max_sched_tunable_scaling = SCHED_TUNABLESCALING_END-1;
 265 #endif
 266
 267 #ifdef CONFIG_COMPACTION
 268 static int min_extfrag_threshold;
 269 static int max_extfrag_threshold = 1000;
 270 #endif
 271
 272 static struct ctl_table kern_table[] = {
 273         {
 274                 .procname       = "sched_child_runs_first",
 275                 .data           = &sysctl_sched_child_runs_first,
 276                 .maxlen         = sizeof(unsigned int),
 277                 .mode           = 0644,
 278                 .proc_handler   = proc_dointvec,
 279         },
 280 #ifdef CONFIG_SCHED_DEBUG
 281         {
 282                 .procname       = "sched_min_granularity_ns",
 283                 .data           = &sysctl_sched_min_granularity,
 284                 .maxlen         = sizeof(unsigned int),
 285                 .mode           = 0644,
 286                 .proc_handler   = sched_proc_update_handler,
 287                 .extra1         = &min_sched_granularity_ns,
 288                 .extra2         = &max_sched_granularity_ns,
 289         },
 290         {
 291                 .procname       = "sched_latency_ns",
 292                 .data           = &sysctl_sched_latency,
 293                 .maxlen         = sizeof(unsigned int),
 294                 .mode           = 0644,
 295                 .proc_handler   = sched_proc_update_handler,
 296                 .extra1         = &min_sched_granularity_ns,
 297                 .extra2         = &max_sched_granularity_ns,
 298         },
 299         {
 300                 .procname       = "sched_wakeup_granularity_ns",
 301                 .data           = &sysctl_sched_wakeup_granularity,
 302                 .maxlen         = sizeof(unsigned int),
 303                 .mode           = 0644,
 304                 .proc_handler   = sched_proc_update_handler,
 305                 .extra1         = &min_wakeup_granularity_ns,
 306                 .extra2         = &max_wakeup_granularity_ns,
 307         },
 308         {
 309                 .procname       = "sched_tunable_scaling",
 310                 .data           = &sysctl_sched_tunable_scaling,
 311                 .maxlen         = sizeof(enum sched_tunable_scaling),
 312                 .mode           = 0644,
 313                 .proc_handler   = sched_proc_update_handler,
 314                 .extra1         = &min_sched_tunable_scaling,
 315                 .extra2         = &max_sched_tunable_scaling,
 316         },
 317         {
 318                 .procname       = "sched_migration_cost",
 319                 .data           = &sysctl_sched_migration_cost,
 320                 .maxlen         = sizeof(unsigned int),
 321                 .mode           = 0644,
 322                 .proc_handler   = proc_dointvec,
 323         },
 324         {
 325                 .procname       = "sched_nr_migrate",
 326                 .data           = &sysctl_sched_nr_migrate,
 327                 .maxlen         = sizeof(unsigned int),
 328                 .mode           = 0644,
 329                 .proc_handler   = proc_dointvec,
 330         },
 331         {
 332                 .procname       = "sched_time_avg",
 333                 .data           = &sysctl_sched_time_avg,
 334                 .maxlen         = sizeof(unsigned int),
 335                 .mode           = 0644,
 336                 .proc_handler   = proc_dointvec,
 337         },
 338         {
 339                 .procname       = "sched_shares_window",
 340                 .data           = &sysctl_sched_shares_window,
 341                 .maxlen         = sizeof(unsigned int),
 342                 .mode           = 0644,
 343                 .proc_handler   = proc_dointvec,
 344         },
 345         {
 346                 .procname       = "timer_migration",
 347                 .data           = &sysctl_timer_migration,
 348                 .maxlen         = sizeof(unsigned int),
 349                 .mode           = 0644,
 350                 .proc_handler   = proc_dointvec_minmax,
 351                 .extra1         = &zero,
 352                 .extra2         = &one,
 353         },
 354 #endif
 355         {
 356                 .procname       = "sched_rt_period_us",
 357                 .data           = &sysctl_sched_rt_period,
 358                 .maxlen         = sizeof(unsigned int),
 359                 .mode           = 0644,
 360                 .proc_handler   = sched_rt_handler,
 361         },
 362         {
 363                 .procname       = "sched_rt_runtime_us",
 364                 .data           = &sysctl_sched_rt_runtime,
 365                 .maxlen         = sizeof(int),
 366                 .mode           = 0644,
 367                 .proc_handler   = sched_rt_handler,
 368         },
 369         {
 370                 .procname       = "sched_compat_yield",
 371                 .data           = &sysctl_sched_compat_yield,
 372                 .maxlen         = sizeof(unsigned int),
 373                 .mode           = 0644,
 374                 .proc_handler   = proc_dointvec,
 375         },
 376 #ifdef CONFIG_SCHED_AUTOGROUP
 377         {
 378                 .procname       = "sched_autogroup_enabled",
 379                 .data           = &sysctl_sched_autogroup_enabled,
 380                 .maxlen         = sizeof(unsigned int),
 381                 .mode           = 0644,
 382                 .proc_handler   = proc_dointvec,
 383                 .extra1         = &zero,
 384                 .extra2         = &one,
 385         },
 386 #endif
 387 #ifdef CONFIG_PROVE_LOCKING
 388         {
 389                 .procname       = "prove_locking",
 390                 .data           = &prove_locking,
 391                 .maxlen         = sizeof(int),
 392                 .mode           = 0644,
 393                 .proc_handler   = proc_dointvec,
 394         },
 395 #endif
 396 #ifdef CONFIG_LOCK_STAT
 397         {
 398                 .procname       = "lock_stat",
 399                 .data           = &lock_stat,
 400                 .maxlen         = sizeof(int),
 401                 .mode           = 0644,
 402                 .proc_handler   = proc_dointvec,
 403         },
 404 #endif
 405         {
 406                 .procname       = "panic",
 407                 .data           = &panic_timeout,
 408                 .maxlen         = sizeof(int),
 409                 .mode           = 0644,
 410                 .proc_handler   = proc_dointvec,
 411         },
 412         {
 413                 .procname       = "core_uses_pid",
 414                 .data           = &core_uses_pid,
 415                 .maxlen         = sizeof(int),
 416                 .mode           = 0644,
 417                 .proc_handler   = proc_dointvec,
 418         },
 419         {
 420                 .procname       = "core_pattern",
 421                 .data           = core_pattern,
 422                 .maxlen         = CORENAME_MAX_SIZE,
 423                 .mode           = 0644,
 424                 .proc_handler   = proc_dostring,
 425         },
 426         {
 427                 .procname       = "core_pipe_limit",
 428                 .data           = &core_pipe_limit,
 429                 .maxlen         = sizeof(unsigned int),
 430                 .mode           = 0644,
 431                 .proc_handler   = proc_dointvec,
 432         },
 433 #ifdef CONFIG_PROC_SYSCTL
 434         {
 435                 .procname       = "tainted",
 436                 .maxlen         = sizeof(long),
 437                 .mode           = 0644,
 438                 .proc_handler   = proc_taint,
 439         },
 440 #endif
 441 #ifdef CONFIG_LATENCYTOP
 442         {
 443                 .procname       = "latencytop",
 444                 .data           = &latencytop_enabled,
 445                 .maxlen         = sizeof(int),
 446                 .mode           = 0644,
 447                 .proc_handler   = proc_dointvec,
 448         },
 449 #endif
 450 #ifdef CONFIG_BLK_DEV_INITRD
 451         {
 452                 .procname       = "real-root-dev",
 453                 .data           = &real_root_dev,
 454                 .maxlen         = sizeof(int),
 455                 .mode           = 0644,
 456                 .proc_handler   = proc_dointvec,
 457         },
 458 #endif
 459         {
 460                 .procname       = "print-fatal-signals",
 461                 .data           = &print_fatal_signals,
 462                 .maxlen         = sizeof(int),
 463                 .mode           = 0644,
 464                 .proc_handler   = proc_dointvec,
 465         },
 466 #ifdef CONFIG_SPARC
 467         {
 468                 .procname       = "reboot-cmd",
 469                 .data           = reboot_command,
 470                 .maxlen         = 256,
 471                 .mode           = 0644,
 472                 .proc_handler   = proc_dostring,
 473         },
 474         {
 475                 .procname       = "stop-a",
 476                 .data           = &stop_a_enabled,
 477                 .maxlen         = sizeof (int),
 478                 .mode           = 0644,
 479                 .proc_handler   = proc_dointvec,
 480         },
 481         {
 482                 .procname       = "scons-poweroff",
 483                 .data           = &scons_pwroff,
 484                 .maxlen         = sizeof (int),
 485                 .mode           = 0644,
 486                 .proc_handler   = proc_dointvec,
 487         },
 488 #endif
 489 #ifdef CONFIG_SPARC64
 490         {
 491                 .procname       = "tsb-ratio",
 492                 .data           = &sysctl_tsb_ratio,
 493                 .maxlen         = sizeof (int),
 494                 .mode           = 0644,
 495                 .proc_handler   = proc_dointvec,
 496         },
 497 #endif
 498 #ifdef __hppa__
 499         {
 500                 .procname       = "soft-power",
 501                 .data           = &pwrsw_enabled,
 502                 .maxlen         = sizeof (int),
 503                 .mode           = 0644,
 504                 .proc_handler   = proc_dointvec,
 505         },
 506         {
 507                 .procname       = "unaligned-trap",
 508                 .data           = &unaligned_enabled,
 509                 .maxlen         = sizeof (int),
 510                 .mode           = 0644,
 511                 .proc_handler   = proc_dointvec,
 512         },
 513 #endif
 514         {
 515                 .procname       = "ctrl-alt-del",
 516                 .data           = &C_A_D,
 517                 .maxlen         = sizeof(int),
 518                 .mode           = 0644,
 519                 .proc_handler   = proc_dointvec,
 520         },
 521 #ifdef CONFIG_FUNCTION_TRACER
 522         {
 523                 .procname       = "ftrace_enabled",
 524                 .data           = &ftrace_enabled,
 525                 .maxlen         = sizeof(int),
 526                 .mode           = 0644,
 527                 .proc_handler   = ftrace_enable_sysctl,
 528         },
 529 #endif
 530 #ifdef CONFIG_STACK_TRACER
 531         {
 532                 .procname       = "stack_tracer_enabled",
 533                 .data           = &stack_tracer_enabled,
 534                 .maxlen         = sizeof(int),
 535                 .mode           = 0644,
 536                 .proc_handler   = stack_trace_sysctl,
 537         },
 538 #endif
 539 #ifdef CONFIG_TRACING
 540         {
 541                 .procname       = "ftrace_dump_on_oops",
 542                 .data           = &ftrace_dump_on_oops,
 543                 .maxlen         = sizeof(int),
 544                 .mode           = 0644,
 545                 .proc_handler   = proc_dointvec,
 546         },
 547 #endif
 548 #ifdef CONFIG_MODULES
 549         {
 550                 .procname       = "modprobe",
 551                 .data           = &modprobe_path,
 552                 .maxlen         = KMOD_PATH_LEN,
 553                 .mode           = 0644,
 554                 .proc_handler   = proc_dostring,
 555         },
 556         {
 557                 .procname       = "modules_disabled",
 558                 .data           = &modules_disabled,
 559                 .maxlen         = sizeof(int),
 560                 .mode           = 0644,
 561                 /* only handle a transition from default "0" to "1" */
 562                 .proc_handler   = proc_dointvec_minmax,
 563                 .extra1         = &one,
 564                 .extra2         = &one,
 565         },
 566 #endif
 567 #ifdef CONFIG_HOTPLUG
 568         {
 569                 .procname       = "hotplug",
 570                 .data           = &uevent_helper,
 571                 .maxlen         = UEVENT_HELPER_PATH_LEN,
 572                 .mode           = 0644,
 573                 .proc_handler   = proc_dostring,
 574         },
 575 #endif
 576 #ifdef CONFIG_CHR_DEV_SG
 577         {
 578                 .procname       = "sg-big-buff",
 579                 .data           = &sg_big_buff,
 580                 .maxlen         = sizeof (int),
 581                 .mode           = 0444,
 582                 .proc_handler   = proc_dointvec,
 583         },
 584 #endif
 585 #ifdef CONFIG_BSD_PROCESS_ACCT
 586         {
 587                 .procname       = "acct",
 588                 .data           = &acct_parm,
 589                 .maxlen         = 3*sizeof(int),
 590                 .mode           = 0644,
 591                 .proc_handler   = proc_dointvec,
 592         },
 593 #endif
 594 #ifdef CONFIG_MAGIC_SYSRQ
 595         {
 596                 .procname       = "sysrq",
 597                 .data           = &__sysrq_enabled,
 598                 .maxlen         = sizeof (int),
 599                 .mode           = 0644,
 600                 .proc_handler   = sysrq_sysctl_handler,
 601         },
 602 #endif
 603 #ifdef CONFIG_PROC_SYSCTL
 604         {
 605                 .procname       = "cad_pid",
 606                 .data           = NULL,
 607                 .maxlen         = sizeof (int),
 608                 .mode           = 0600,
 609                 .proc_handler   = proc_do_cad_pid,
 610         },
 611 #endif
 612         {
 613                 .procname       = "threads-max",
 614                 .data           = &max_threads,
 615                 .maxlen         = sizeof(int),
 616                 .mode           = 0644,
 617                 .proc_handler   = proc_dointvec,
 618         },
 619         {
 620                 .procname       = "random",
 621                 .mode           = 0555,
 622                 .child          = random_table,
 623         },
 624         {
 625                 .procname       = "overflowuid",
 626                 .data           = &overflowuid,
 627                 .maxlen         = sizeof(int),
 628                 .mode           = 0644,
 629                 .proc_handler   = proc_dointvec_minmax,
 630                 .extra1         = &minolduid,
 631                 .extra2         = &maxolduid,
 632         },
 633         {
 634                 .procname       = "overflowgid",
 635                 .data           = &overflowgid,
 636                 .maxlen         = sizeof(int),
 637                 .mode           = 0644,
 638                 .proc_handler   = proc_dointvec_minmax,
 639                 .extra1         = &minolduid,
 640                 .extra2         = &maxolduid,
 641         },
 642 #ifdef CONFIG_S390
 643 #ifdef CONFIG_MATHEMU
 644         {
 645                 .procname       = "ieee_emulation_warnings",
 646                 .data           = &sysctl_ieee_emulation_warnings,
 647                 .maxlen         = sizeof(int),
 648                 .mode           = 0644,
 649                 .proc_handler   = proc_dointvec,
 650         },
 651 #endif
 652         {
 653                 .procname       = "userprocess_debug",
 654                 .data           = &show_unhandled_signals,
 655                 .maxlen         = sizeof(int),
 656                 .mode           = 0644,
 657                 .proc_handler   = proc_dointvec,
 658         },
 659 #endif
 660         {
 661                 .procname       = "pid_max",
 662                 .data           = &pid_max,
 663                 .maxlen         = sizeof (int),
 664                 .mode           = 0644,
 665                 .proc_handler   = proc_dointvec_minmax,
 666                 .extra1         = &pid_max_min,
 667                 .extra2         = &pid_max_max,
 668         },
 669         {
 670                 .procname       = "panic_on_oops",
 671                 .data           = &panic_on_oops,
 672                 .maxlen         = sizeof(int),
 673                 .mode           = 0644,
 674                 .proc_handler   = proc_dointvec,
 675         },
 676 #if defined CONFIG_PRINTK
 677         {
 678                 .procname       = "printk",
 679                 .data           = &console_loglevel,
 680                 .maxlen         = 4*sizeof(int),
 681                 .mode           = 0644,
 682                 .proc_handler   = proc_dointvec,
 683         },
 684         {
 685                 .procname       = "printk_ratelimit",
 686                 .data           = &printk_ratelimit_state.interval,
 687                 .maxlen         = sizeof(int),
 688                 .mode           = 0644,
 689                 .proc_handler   = proc_dointvec_jiffies,
 690         },
 691         {
 692                 .procname       = "printk_ratelimit_burst",
 693                 .data           = &printk_ratelimit_state.burst,
 694                 .maxlen         = sizeof(int),
 695                 .mode           = 0644,
 696                 .proc_handler   = proc_dointvec,
 697         },
 698         {
 699                 .procname       = "printk_delay",
 700                 .data           = &printk_delay_msec,
 701                 .maxlen         = sizeof(int),
 702                 .mode           = 0644,
 703                 .proc_handler   = proc_dointvec_minmax,
 704                 .extra1         = &zero,
 705                 .extra2         = &ten_thousand,
 706         },
 707         {
 708                 .procname       = "dmesg_restrict",
 709                 .data           = &dmesg_restrict,
 710                 .maxlen         = sizeof(int),
 711                 .mode           = 0644,
 712                 .proc_handler   = proc_dointvec_minmax,
 713                 .extra1         = &zero,
 714                 .extra2         = &one,
 715         },
 716         {
 717                 .procname       = "kptr_restrict",
 718                 .data           = &kptr_restrict,
 719                 .maxlen         = sizeof(int),
 720                 .mode           = 0644,
 721                 .proc_handler   = proc_dmesg_restrict,
 722                 .extra1         = &zero,
 723                 .extra2         = &two,
 724         },
 725 #endif
 726         {
 727                 .procname       = "ngroups_max",
 728                 .data           = &ngroups_max,
 729                 .maxlen         = sizeof (int),
 730                 .mode           = 0444,
 731                 .proc_handler   = proc_dointvec,
 732         },
 733 #if defined(CONFIG_LOCKUP_DETECTOR)
 734         {
 735                 .procname       = "watchdog",
 736                 .data           = &watchdog_enabled,
 737                 .maxlen         = sizeof (int),
 738                 .mode           = 0644,
 739                 .proc_handler   = proc_dowatchdog_enabled,
 740         },
 741         {
 742                 .procname       = "watchdog_thresh",
 743                 .data           = &softlockup_thresh,
 744                 .maxlen         = sizeof(int),
 745                 .mode           = 0644,
 746                 .proc_handler   = proc_dowatchdog_thresh,
 747                 .extra1         = &neg_one,
 748                 .extra2         = &sixty,
 749         },
 750         {
 751                 .procname       = "softlockup_panic",
 752                 .data           = &softlockup_panic,
 753                 .maxlen         = sizeof(int),
 754                 .mode           = 0644,
 755                 .proc_handler   = proc_dointvec_minmax,
 756                 .extra1         = &zero,
 757                 .extra2         = &one,
 758         },
 759         {
 760                 .procname       = "nmi_watchdog",
 761                 .data           = &watchdog_enabled,
 762                 .maxlen         = sizeof (int),
 763                 .mode           = 0644,
 764                 .proc_handler   = proc_dowatchdog_enabled,
 765         },
 766 #endif
 767 #if defined(CONFIG_X86_LOCAL_APIC) && defined(CONFIG_X86)
 768         {
 769                 .procname       = "unknown_nmi_panic",
 770                 .data           = &unknown_nmi_panic,
 771                 .maxlen         = sizeof (int),
 772                 .mode           = 0644,
 773                 .proc_handler   = proc_dointvec,
 774         },
 775 #endif
 776 #if defined(CONFIG_X86)
 777         {
 778                 .procname       = "panic_on_unrecovered_nmi",
 779                 .data           = &panic_on_unrecovered_nmi,
 780                 .maxlen         = sizeof(int),
 781                 .mode           = 0644,
 782                 .proc_handler   = proc_dointvec,
 783         },
 784         {
 785                 .procname       = "panic_on_io_nmi",
 786                 .data           = &panic_on_io_nmi,
 787                 .maxlen         = sizeof(int),
 788                 .mode           = 0644,
 789                 .proc_handler   = proc_dointvec,
 790         },
 791         {
 792                 .procname       = "bootloader_type",
 793                 .data           = &bootloader_type,
 794                 .maxlen         = sizeof (int),
 795                 .mode           = 0444,
 796                 .proc_handler   = proc_dointvec,
 797         },
 798         {
 799                 .procname       = "bootloader_version",
 800                 .data           = &bootloader_version,
 801                 .maxlen         = sizeof (int),
 802                 .mode           = 0444,
 803                 .proc_handler   = proc_dointvec,
 804         },
 805         {
 806                 .procname       = "kstack_depth_to_print",
 807                 .data           = &kstack_depth_to_print,
 808                 .maxlen         = sizeof(int),
 809                 .mode           = 0644,
 810                 .proc_handler   = proc_dointvec,
 811         },
 812         {
 813                 .procname       = "io_delay_type",
 814                 .data           = &io_delay_type,
 815                 .maxlen         = sizeof(int),
 816                 .mode           = 0644,
 817                 .proc_handler   = proc_dointvec,
 818         },
 819 #endif
 820 #if defined(CONFIG_MMU)
 821         {
 822                 .procname       = "randomize_va_space",
 823                 .data           = &randomize_va_space,
 824                 .maxlen         = sizeof(int),
 825                 .mode           = 0644,
 826                 .proc_handler   = proc_dointvec,
 827         },
 828 #endif
 829 #if defined(CONFIG_S390) && defined(CONFIG_SMP)
 830         {
 831                 .procname       = "spin_retry",
 832                 .data           = &spin_retry,
 833                 .maxlen         = sizeof (int),
 834                 .mode           = 0644,
 835                 .proc_handler   = proc_dointvec,
 836         },
 837 #endif
 838 #if     defined(CONFIG_ACPI_SLEEP) && defined(CONFIG_X86)
 839         {
 840                 .procname       = "acpi_video_flags",
 841                 .data           = &acpi_realmode_flags,
 842                 .maxlen         = sizeof (unsigned long),
 843                 .mode           = 0644,
 844                 .proc_handler   = proc_doulongvec_minmax,
 845         },
 846 #endif
 847 #ifdef CONFIG_IA64
 848         {
 849                 .procname       = "ignore-unaligned-usertrap",
 850                 .data           = &no_unaligned_warning,
 851                 .maxlen         = sizeof (int),
 852                 .mode           = 0644,
 853                 .proc_handler   = proc_dointvec,
 854         },
 855         {
 856                 .procname       = "unaligned-dump-stack",
 857                 .data           = &unaligned_dump_stack,
 858                 .maxlen         = sizeof (int),
 859                 .mode           = 0644,
 860                 .proc_handler   = proc_dointvec,
 861         },
 862 #endif
 863 #ifdef CONFIG_DETECT_HUNG_TASK
 864         {
 865                 .procname       = "hung_task_panic",
 866                 .data           = &sysctl_hung_task_panic,
 867                 .maxlen         = sizeof(int),
 868                 .mode           = 0644,
 869                 .proc_handler   = proc_dointvec_minmax,
 870                 .extra1         = &zero,
 871                 .extra2         = &one,
 872         },
 873         {
 874                 .procname       = "hung_task_check_count",
 875                 .data           = &sysctl_hung_task_check_count,
 876                 .maxlen         = sizeof(unsigned long),
 877                 .mode           = 0644,
 878                 .proc_handler   = proc_doulongvec_minmax,
 879         },
 880         {
 881                 .procname       = "hung_task_timeout_secs",
 882                 .data           = &sysctl_hung_task_timeout_secs,
 883                 .maxlen         = sizeof(unsigned long),
 884                 .mode           = 0644,
 885                 .proc_handler   = proc_dohung_task_timeout_secs,
 886         },
 887         {
 888                 .procname       = "hung_task_warnings",
 889                 .data           = &sysctl_hung_task_warnings,
 890                 .maxlen         = sizeof(unsigned long),
 891                 .mode           = 0644,
 892                 .proc_handler   = proc_doulongvec_minmax,
 893         },
 894 #endif
 895 #ifdef CONFIG_COMPAT
 896         {
 897                 .procname       = "compat-log",
 898                 .data           = &compat_log,
 899                 .maxlen         = sizeof (int),
 900                 .mode           = 0644,
 901                 .proc_handler   = proc_dointvec,
 902         },
 903 #endif
 904 #ifdef CONFIG_RT_MUTEXES
 905         {
 906                 .procname       = "max_lock_depth",
 907                 .data           = &max_lock_depth,
 908                 .maxlen         = sizeof(int),
 909                 .mode           = 0644,
 910                 .proc_handler   = proc_dointvec,
 911         },
 912 #endif
 913         {
 914                 .procname       = "poweroff_cmd",
 915                 .data           = &poweroff_cmd,
 916                 .maxlen         = POWEROFF_CMD_PATH_LEN,
 917                 .mode           = 0644,
 918                 .proc_handler   = proc_dostring,
 919         },
 920 #ifdef CONFIG_KEYS
 921         {
 922                 .procname       = "keys",
 923                 .mode           = 0555,
 924                 .child          = key_sysctls,
 925         },
 926 #endif
 927 #ifdef CONFIG_RCU_TORTURE_TEST
 928         {
 929                 .procname       = "rcutorture_runnable",
 930                 .data           = &rcutorture_runnable,
 931                 .maxlen         = sizeof(int),
 932                 .mode           = 0644,
 933                 .proc_handler   = proc_dointvec,
 934         },
 935 #endif
 936 #ifdef CONFIG_PERF_EVENTS
 937         {
 938                 .procname       = "perf_event_paranoid",
 939                 .data           = &sysctl_perf_event_paranoid,
 940                 .maxlen         = sizeof(sysctl_perf_event_paranoid),
 941                 .mode           = 0644,
 942                 .proc_handler   = proc_dointvec,
 943         },
 944         {
 945                 .procname       = "perf_event_mlock_kb",
 946                 .data           = &sysctl_perf_event_mlock,
 947                 .maxlen         = sizeof(sysctl_perf_event_mlock),
 948                 .mode           = 0644,
 949                 .proc_handler   = proc_dointvec,
 950         },
 951         {
 952                 .procname       = "perf_event_max_sample_rate",
 953                 .data           = &sysctl_perf_event_sample_rate,
 954                 .maxlen         = sizeof(sysctl_perf_event_sample_rate),
 955                 .mode           = 0644,
 956                 .proc_handler   = proc_dointvec,
 957         },
 958 #endif
 959 #ifdef CONFIG_KMEMCHECK
 960         {
 961                 .procname       = "kmemcheck",
 962                 .data           = &kmemcheck_enabled,
 963                 .maxlen         = sizeof(int),
 964                 .mode           = 0644,
 965                 .proc_handler   = proc_dointvec,
 966         },
 967 #endif
 968 #ifdef CONFIG_BLOCK
 969         {
 970                 .procname       = "blk_iopoll",
 971                 .data           = &blk_iopoll_enabled,
 972                 .maxlen         = sizeof(int),
 973                 .mode           = 0644,
 974                 .proc_handler   = proc_dointvec,
 975         },
 976 #endif
 977         { }
 978 };
 979
 980 static struct ctl_table vm_table[] = {
 981         {
 982                 .procname       = "overcommit_memory",
 983                 .data           = &sysctl_overcommit_memory,
 984                 .maxlen         = sizeof(sysctl_overcommit_memory),
 985                 .mode           = 0644,
 986                 .proc_handler   = proc_dointvec,
 987         },
 988         {
 989                 .procname       = "panic_on_oom",
 990                 .data           = &sysctl_panic_on_oom,
 991                 .maxlen         = sizeof(sysctl_panic_on_oom),
 992                 .mode           = 0644,
 993                 .proc_handler   = proc_dointvec,
 994         },
 995         {
 996                 .procname       = "oom_kill_allocating_task",
 997                 .data           = &sysctl_oom_kill_allocating_task,
 998                 .maxlen         = sizeof(sysctl_oom_kill_allocating_task),
 999                 .mode           = 0644,
1000                 .proc_handler   = proc_dointvec,
1001         },
1002         {
1003                 .procname       = "oom_dump_tasks",
1004                 .data           = &sysctl_oom_dump_tasks,
1005                 .maxlen         = sizeof(sysctl_oom_dump_tasks),
1006                 .mode           = 0644,
1007                 .proc_handler   = proc_dointvec,
1008         },
1009         {
1010                 .procname       = "overcommit_ratio",
1011                 .data           = &sysctl_overcommit_ratio,
1012                 .maxlen         = sizeof(sysctl_overcommit_ratio),
1013                 .mode           = 0644,
1014                 .proc_handler   = proc_dointvec,
1015         },
1016         {
1017                 .procname       = "page-cluster",
1018                 .data           = &page_cluster,
1019                 .maxlen         = sizeof(int),
1020                 .mode           = 0644,
1021                 .proc_handler   = proc_dointvec,
1022         },
1023         {
1024                 .procname       = "dirty_background_ratio",
1025                 .data           = &dirty_background_ratio,
1026                 .maxlen         = sizeof(dirty_background_ratio),
1027                 .mode           = 0644,
1028                 .proc_handler   = dirty_background_ratio_handler,
1029                 .extra1         = &zero,
1030                 .extra2         = &one_hundred,
1031         },
1032         {
1033                 .procname       = "dirty_background_bytes",
1034                 .data           = &dirty_background_bytes,
1035                 .maxlen         = sizeof(dirty_background_bytes),
1036                 .mode           = 0644,
1037                 .proc_handler   = dirty_background_bytes_handler,
1038                 .extra1         = &one_ul,
1039         },
1040         {
1041                 .procname       = "dirty_ratio",
1042                 .data           = &vm_dirty_ratio,
1043                 .maxlen         = sizeof(vm_dirty_ratio),
1044                 .mode           = 0644,
1045                 .proc_handler   = dirty_ratio_handler,
1046                 .extra1         = &zero,
1047                 .extra2         = &one_hundred,
1048         },
1049         {
1050                 .procname       = "dirty_bytes",
1051                 .data           = &vm_dirty_bytes,
1052                 .maxlen         = sizeof(vm_dirty_bytes),
1053                 .mode           = 0644,
1054                 .proc_handler   = dirty_bytes_handler,
1055                 .extra1         = &dirty_bytes_min,
1056         },
1057         {
1058                 .procname       = "dirty_writeback_centisecs",
1059                 .data           = &dirty_writeback_interval,
1060                 .maxlen         = sizeof(dirty_writeback_interval),
1061                 .mode           = 0644,
1062                 .proc_handler   = dirty_writeback_centisecs_handler,
1063         },
1064         {
1065                 .procname       = "dirty_expire_centisecs",
1066                 .data           = &dirty_expire_interval,
1067                 .maxlen         = sizeof(dirty_expire_interval),
1068                 .mode           = 0644,
1069                 .proc_handler   = proc_dointvec,
1070         },
1071         {
1072                 .procname       = "nr_pdflush_threads",
1073                 .data           = &nr_pdflush_threads,
1074                 .maxlen         = sizeof nr_pdflush_threads,
1075                 .mode           = 0444 /* read-only*/,
1076                 .proc_handler   = proc_dointvec,
1077         },
1078         {
1079                 .procname       = "swappiness",
1080                 .data           = &vm_swappiness,
1081                 .maxlen         = sizeof(vm_swappiness),
1082                 .mode           = 0644,
1083                 .proc_handler   = proc_dointvec_minmax,
1084                 .extra1         = &zero,
1085                 .extra2         = &one_hundred,
1086         },
1087 #ifdef CONFIG_HUGETLB_PAGE
1088         {
1089                 .procname       = "nr_hugepages",
1090                 .data           = NULL,
1091                 .maxlen         = sizeof(unsigned long),
1092                 .mode           = 0644,
1093                 .proc_handler   = hugetlb_sysctl_handler,
1094                 .extra1         = (void *)&hugetlb_zero,
1095                 .extra2         = (void *)&hugetlb_infinity,
1096         },
1097 #ifdef CONFIG_NUMA
1098         {
1099                 .procname       = "nr_hugepages_mempolicy",
1100                 .data           = NULL,
1101                 .maxlen         = sizeof(unsigned long),
1102                 .mode           = 0644,
1103                 .proc_handler   = &hugetlb_mempolicy_sysctl_handler,
1104                 .extra1         = (void *)&hugetlb_zero,
1105                 .extra2         = (void *)&hugetlb_infinity,
1106         },
1107 #endif
1108          {
1109                 .procname       = "hugetlb_shm_group",
1110                 .data           = &sysctl_hugetlb_shm_group,
1111                 .maxlen         = sizeof(gid_t),
1112                 .mode           = 0644,
1113                 .proc_handler   = proc_dointvec,
1114          },
1115          {
1116                 .procname       = "hugepages_treat_as_movable",
1117                 .data           = &hugepages_treat_as_movable,
1118                 .maxlen         = sizeof(int),
1119                 .mode           = 0644,
1120                 .proc_handler   = hugetlb_treat_movable_handler,
1121         },
1122         {
1123                 .procname       = "nr_overcommit_hugepages",
1124                 .data           = NULL,
1125                 .maxlen         = sizeof(unsigned long),
1126                 .mode           = 0644,
1127                 .proc_handler   = hugetlb_overcommit_handler,
1128                 .extra1         = (void *)&hugetlb_zero,
1129                 .extra2         = (void *)&hugetlb_infinity,
1130         },
1131 #endif
1132         {
1133                 .procname       = "lowmem_reserve_ratio",
1134                 .data           = &sysctl_lowmem_reserve_ratio,
1135                 .maxlen         = sizeof(sysctl_lowmem_reserve_ratio),
1136                 .mode           = 0644,
1137                 .proc_handler   = lowmem_reserve_ratio_sysctl_handler,
1138         },
1139         {
1140                 .procname       = "drop_caches",
1141                 .data           = &sysctl_drop_caches,
1142                 .maxlen         = sizeof(int),
1143                 .mode           = 0644,
1144                 .proc_handler   = drop_caches_sysctl_handler,
1145         },
1146 #ifdef CONFIG_COMPACTION
1147         {
1148                 .procname       = "compact_memory",
1149                 .data           = &sysctl_compact_memory,
1150                 .maxlen         = sizeof(int),
1151                 .mode           = 0200,
1152                 .proc_handler   = sysctl_compaction_handler,
1153         },
1154         {
1155                 .procname       = "extfrag_threshold",
1156                 .data           = &sysctl_extfrag_threshold,
1157                 .maxlen         = sizeof(int),
1158                 .mode           = 0644,
1159                 .proc_handler   = sysctl_extfrag_handler,
1160                 .extra1         = &min_extfrag_threshold,
1161                 .extra2         = &max_extfrag_threshold,
1162         },
1163
1164 #endif /* CONFIG_COMPACTION */
1165         {
1166                 .procname       = "min_free_kbytes",
1167                 .data           = &min_free_kbytes,
1168                 .maxlen         = sizeof(min_free_kbytes),
1169                 .mode           = 0644,
1170                 .proc_handler   = min_free_kbytes_sysctl_handler,
1171                 .extra1         = &zero,
1172         },
1173         {
1174                 .procname       = "percpu_pagelist_fraction",
1175                 .data           = &percpu_pagelist_fraction,
1176                 .maxlen         = sizeof(percpu_pagelist_fraction),
1177                 .mode           = 0644,
1178                 .proc_handler   = percpu_pagelist_fraction_sysctl_handler,
1179                 .extra1         = &min_percpu_pagelist_fract,
1180         },
1181 #ifdef CONFIG_MMU
1182         {
1183                 .procname       = "max_map_count",
1184                 .data           = &sysctl_max_map_count,
1185                 .maxlen         = sizeof(sysctl_max_map_count),
1186                 .mode           = 0644,
1187                 .proc_handler   = proc_dointvec_minmax,
1188                 .extra1         = &zero,
1189         },
1190 #else
1191         {
1192                 .procname       = "nr_trim_pages",
1193                 .data           = &sysctl_nr_trim_pages,
1194                 .maxlen         = sizeof(sysctl_nr_trim_pages),
1195                 .mode           = 0644,
1196                 .proc_handler   = proc_dointvec_minmax,
1197                 .extra1         = &zero,
1198         },
1199 #endif
1200         {
1201                 .procname       = "laptop_mode",
1202                 .data           = &laptop_mode,
1203                 .maxlen         = sizeof(laptop_mode),
1204                 .mode           = 0644,
1205                 .proc_handler   = proc_dointvec_jiffies,
1206         },
1207         {
1208                 .procname       = "block_dump",
1209                 .data           = &block_dump,
1210                 .maxlen         = sizeof(block_dump),
1211                 .mode           = 0644,
1212                 .proc_handler   = proc_dointvec,
1213                 .extra1         = &zero,
1214         },
1215         {
1216                 .procname       = "vfs_cache_pressure",
1217                 .data           = &sysctl_vfs_cache_pressure,
1218                 .maxlen         = sizeof(sysctl_vfs_cache_pressure),
1219                 .mode           = 0644,
1220                 .proc_handler   = proc_dointvec,
1221                 .extra1         = &zero,
1222         },
1223 #ifdef HAVE_ARCH_PICK_MMAP_LAYOUT
1224         {
1225                 .procname       = "legacy_va_layout",
1226                 .data           = &sysctl_legacy_va_layout,
1227                 .maxlen         = sizeof(sysctl_legacy_va_layout),
1228                 .mode           = 0644,
1229                 .proc_handler   = proc_dointvec,
1230                 .extra1         = &zero,
1231         },
1232 #endif
1233 #ifdef CONFIG_NUMA
1234         {
1235                 .procname       = "zone_reclaim_mode",
1236                 .data           = &zone_reclaim_mode,
1237                 .maxlen         = sizeof(zone_reclaim_mode),
1238                 .mode           = 0644,
1239                 .proc_handler   = proc_dointvec,
1240                 .extra1         = &zero,
1241         },
1242         {
1243                 .procname       = "min_unmapped_ratio",
1244                 .data           = &sysctl_min_unmapped_ratio,
1245                 .maxlen         = sizeof(sysctl_min_unmapped_ratio),
1246                 .mode           = 0644,
1247                 .proc_handler   = sysctl_min_unmapped_ratio_sysctl_handler,
1248                 .extra1         = &zero,
1249                 .extra2         = &one_hundred,
1250         },
1251         {
1252                 .procname       = "min_slab_ratio",
1253                 .data           = &sysctl_min_slab_ratio,
1254                 .maxlen         = sizeof(sysctl_min_slab_ratio),
1255                 .mode           = 0644,
1256                 .proc_handler   = sysctl_min_slab_ratio_sysctl_handler,
1257                 .extra1         = &zero,
1258                 .extra2         = &one_hundred,
1259         },
1260 #endif
1261 #ifdef CONFIG_SMP
1262         {
1263                 .procname       = "stat_interval",
1264                 .data           = &sysctl_stat_interval,
1265                 .maxlen         = sizeof(sysctl_stat_interval),
1266                 .mode           = 0644,
1267                 .proc_handler   = proc_dointvec_jiffies,
1268         },
1269 #endif
1270 #ifdef CONFIG_MMU
1271         {
1272                 .procname       = "mmap_min_addr",
1273                 .data           = &dac_mmap_min_addr,
1274                 .maxlen         = sizeof(unsigned long),
1275                 .mode           = 0644,
1276                 .proc_handler   = mmap_min_addr_handler,
1277         },
1278 #endif
1279 #ifdef CONFIG_NUMA
1280         {
1281                 .procname       = "numa_zonelist_order",
1282                 .data           = &numa_zonelist_order,
1283                 .maxlen         = NUMA_ZONELIST_ORDER_LEN,
1284                 .mode           = 0644,
1285                 .proc_handler   = numa_zonelist_order_handler,
1286         },
1287 #endif
1288 #if (defined(CONFIG_X86_32) && !defined(CONFIG_UML))|| \
1289    (defined(CONFIG_SUPERH) && defined(CONFIG_VSYSCALL))
1290         {
1291                 .procname       = "vdso_enabled",
1292                 .data           = &vdso_enabled,
1293                 .maxlen         = sizeof(vdso_enabled),
1294                 .mode           = 0644,
1295                 .proc_handler   = proc_dointvec,
1296                 .extra1         = &zero,
1297         },
1298 #endif
1299 #ifdef CONFIG_HIGHMEM
1300         {
1301                 .procname       = "highmem_is_dirtyable",
1302                 .data           = &vm_highmem_is_dirtyable,
1303                 .maxlen         = sizeof(vm_highmem_is_dirtyable),
1304                 .mode           = 0644,
1305                 .proc_handler   = proc_dointvec_minmax,
1306                 .extra1         = &zero,
1307                 .extra2         = &one,
1308         },
1309 #endif
1310         {
1311                 .procname       = "scan_unevictable_pages",
1312                 .data           = &scan_unevictable_pages,
1313                 .maxlen         = sizeof(scan_unevictable_pages),
1314                 .mode           = 0644,
1315                 .proc_handler   = scan_unevictable_handler,
1316         },
1317 #ifdef CONFIG_MEMORY_FAILURE
1318         {
1319                 .procname       = "memory_failure_early_kill",
1320                 .data           = &sysctl_memory_failure_early_kill,
1321                 .maxlen         = sizeof(sysctl_memory_failure_early_kill),
1322                 .mode           = 0644,
1323                 .proc_handler   = proc_dointvec_minmax,
1324                 .extra1         = &zero,
1325                 .extra2         = &one,
1326         },
1327         {
1328                 .procname       = "memory_failure_recovery",
1329                 .data           = &sysctl_memory_failure_recovery,
1330                 .maxlen         = sizeof(sysctl_memory_failure_recovery),
1331                 .mode           = 0644,
1332                 .proc_handler   = proc_dointvec_minmax,
1333                 .extra1         = &zero,
1334                 .extra2         = &one,
1335         },
1336 #endif
1337         { }
1338 };
1339
1340 #if defined(CONFIG_BINFMT_MISC) || defined(CONFIG_BINFMT_MISC_MODULE)
1341 static struct ctl_table binfmt_misc_table[] = {
1342         { }
1343 };
1344 #endif
1345
1346 static struct ctl_table fs_table[] = {
1347         {
1348                 .procname       = "inode-nr",
1349                 .data           = &inodes_stat,
1350                 .maxlen         = 2*sizeof(int),
1351                 .mode           = 0444,
1352                 .proc_handler   = proc_nr_inodes,
1353         },
1354         {
1355                 .procname       = "inode-state",
1356                 .data           = &inodes_stat,
1357                 .maxlen         = 7*sizeof(int),
1358                 .mode           = 0444,
1359                 .proc_handler   = proc_nr_inodes,
1360         },
1361         {
1362                 .procname       = "file-nr",
1363                 .data           = &files_stat,
1364                 .maxlen         = sizeof(files_stat),
1365                 .mode           = 0444,
1366                 .proc_handler   = proc_nr_files,
1367         },
1368         {
1369                 .procname       = "file-max",
1370                 .data           = &files_stat.max_files,
1371                 .maxlen         = sizeof(files_stat.max_files),
1372                 .mode           = 0644,
1373                 .proc_handler   = proc_doulongvec_minmax,
1374         },
1375         {
1376                 .procname       = "nr_open",
1377                 .data           = &sysctl_nr_open,
1378                 .maxlen         = sizeof(int),
1379                 .mode           = 0644,
1380                 .proc_handler   = proc_dointvec_minmax,
1381                 .extra1         = &sysctl_nr_open_min,
1382                 .extra2         = &sysctl_nr_open_max,
1383         },
1384         {
1385                 .procname       = "dentry-state",
1386                 .data           = &dentry_stat,
1387                 .maxlen         = 6*sizeof(int),
1388                 .mode           = 0444,
1389                 .proc_handler   = proc_nr_dentry,
1390         },
1391         {
1392                 .procname       = "overflowuid",
1393                 .data           = &fs_overflowuid,
1394                 .maxlen         = sizeof(int),
1395                 .mode           = 0644,
1396                 .proc_handler   = proc_dointvec_minmax,
1397                 .extra1         = &minolduid,
1398                 .extra2         = &maxolduid,
1399         },
1400         {
1401                 .procname       = "overflowgid",
1402                 .data           = &fs_overflowgid,
1403                 .maxlen         = sizeof(int),
1404                 .mode           = 0644,
1405                 .proc_handler   = proc_dointvec_minmax,
1406                 .extra1         = &minolduid,
1407                 .extra2         = &maxolduid,
1408         },
1409 #ifdef CONFIG_FILE_LOCKING
1410         {
1411                 .procname       = "leases-enable",
1412                 .data           = &leases_enable,
1413                 .maxlen         = sizeof(int),
1414                 .mode           = 0644,
1415                 .proc_handler   = proc_dointvec,
1416         },
1417 #endif
1418 #ifdef CONFIG_DNOTIFY
1419         {
1420                 .procname       = "dir-notify-enable",
1421                 .data           = &dir_notify_enable,
1422                 .maxlen         = sizeof(int),
1423                 .mode           = 0644,
1424                 .proc_handler   = proc_dointvec,
1425         },
1426 #endif
1427 #ifdef CONFIG_MMU
1428 #ifdef CONFIG_FILE_LOCKING
1429         {
1430                 .procname       = "lease-break-time",
1431                 .data           = &lease_break_time,
1432                 .maxlen         = sizeof(int),
1433                 .mode           = 0644,
1434                 .proc_handler   = proc_dointvec,
1435         },
1436 #endif
1437 #ifdef CONFIG_AIO
1438         {
1439                 .procname       = "aio-nr",
1440                 .data           = &aio_nr,
1441                 .maxlen         = sizeof(aio_nr),
1442                 .mode           = 0444,
1443                 .proc_handler   = proc_doulongvec_minmax,
1444         },
1445         {
1446                 .procname       = "aio-max-nr",
1447                 .data           = &aio_max_nr,
1448                 .maxlen         = sizeof(aio_max_nr),
1449                 .mode           = 0644,
1450                 .proc_handler   = proc_doulongvec_minmax,
1451         },
1452 #endif /* CONFIG_AIO */
1453 #ifdef CONFIG_INOTIFY_USER
1454         {
1455                 .procname       = "inotify",
1456                 .mode           = 0555,
1457                 .child          = inotify_table,
1458         },
1459 #endif
1460 #ifdef CONFIG_EPOLL
1461         {
1462                 .procname       = "epoll",
1463                 .mode           = 0555,
1464                 .child          = epoll_table,
1465         },
1466 #endif
1467 #endif
1468         {
1469                 .procname       = "suid_dumpable",
1470                 .data           = &suid_dumpable,
1471                 .maxlen         = sizeof(int),
1472                 .mode           = 0644,
1473                 .proc_handler   = proc_dointvec_minmax,
1474                 .extra1         = &zero,
1475                 .extra2         = &two,
1476         },
1477 #if defined(CONFIG_BINFMT_MISC) || defined(CONFIG_BINFMT_MISC_MODULE)
1478         {
1479                 .procname       = "binfmt_misc",
1480                 .mode           = 0555,
1481                 .child          = binfmt_misc_table,
1482         },
1483 #endif
1484         {
1485                 .procname       = "pipe-max-size",
1486                 .data           = &pipe_max_size,
1487                 .maxlen         = sizeof(int),
1488                 .mode           = 0644,
1489                 .proc_handler   = &pipe_proc_fn,
1490                 .extra1         = &pipe_min_size,
1491         },
1492         { }
1493 };
1494
1495 static struct ctl_table debug_table[] = {
1496 #if defined(CONFIG_X86) || defined(CONFIG_PPC) || defined(CONFIG_SPARC) || \
1497     defined(CONFIG_S390)
1498         {
1499                 .procname       = "exception-trace",
1500                 .data           = &show_unhandled_signals,
1501                 .maxlen         = sizeof(int),
1502                 .mode           = 0644,
1503                 .proc_handler   = proc_dointvec
1504         },
1505 #endif
1506 #if defined(CONFIG_OPTPROBES)
1507         {
1508                 .procname       = "kprobes-optimization",
1509                 .data           = &sysctl_kprobes_optimization,
1510                 .maxlen         = sizeof(int),
1511                 .mode           = 0644,
1512                 .proc_handler   = proc_kprobes_optimization_handler,
1513                 .extra1         = &zero,
1514                 .extra2         = &one,
1515         },
1516 #endif
1517         { }
1518 };
1519
1520 static struct ctl_table dev_table[] = {
1521         { }
1522 };
1523
1524 static DEFINE_SPINLOCK(sysctl_lock);
1525
1526 /* called under sysctl_lock */
1527 static int use_table(struct ctl_table_header *p)
1528 {
1529         if (unlikely(p->unregistering))
1530                 return 0;
1531         p->used++;
1532         return 1;
1533 }
1534
1535 /* called under sysctl_lock */
1536 static void unuse_table(struct ctl_table_header *p)
1537 {
1538         if (!--p->used)
1539                 if (unlikely(p->unregistering))
1540                         complete(p->unregistering);
1541 }
1542
1543 /* called under sysctl_lock, will reacquire if has to wait */
1544 static void start_unregistering(struct ctl_table_header *p)
1545 {
1546         /*
1547          * if p->used is 0, nobody will ever touch that entry again;
1548          * we'll eliminate all paths to it before dropping sysctl_lock
1549          */
1550         if (unlikely(p->used)) {
1551                 struct completion wait;
1552                 init_completion(&wait);
1553                 p->unregistering = &wait;
1554                 spin_unlock(&sysctl_lock);
1555                 wait_for_completion(&wait);
1556                 spin_lock(&sysctl_lock);
1557         } else {
1558                 /* anything non-NULL; we'll never dereference it */
1559                 p->unregistering = ERR_PTR(-EINVAL);
1560         }
1561         /*
1562          * do not remove from the list until nobody holds it; walking the
1563          * list in do_sysctl() relies on that.
1564          */
1565         list_del_init(&p->ctl_entry);
1566 }
1567
1568 void sysctl_head_get(struct ctl_table_header *head)
1569 {
1570         spin_lock(&sysctl_lock);
1571         head->count++;
1572         spin_unlock(&sysctl_lock);
1573 }
1574
1575 static void free_head(struct rcu_head *rcu)
1576 {
1577         kfree(container_of(rcu, struct ctl_table_header, rcu));
1578 }
1579
1580 void sysctl_head_put(struct ctl_table_header *head)
1581 {
1582         spin_lock(&sysctl_lock);
1583         if (!--head->count)
1584                 call_rcu(&head->rcu, free_head);
1585         spin_unlock(&sysctl_lock);
1586 }
1587
1588 struct ctl_table_header *sysctl_head_grab(struct ctl_table_header *head)
1589 {
1590         if (!head)
1591                 BUG();
1592         spin_lock(&sysctl_lock);
1593         if (!use_table(head))
1594                 head = ERR_PTR(-ENOENT);
1595         spin_unlock(&sysctl_lock);
1596         return head;
1597 }
1598
1599 void sysctl_head_finish(struct ctl_table_header *head)
1600 {
1601         if (!head)
1602                 return;
1603         spin_lock(&sysctl_lock);
1604         unuse_table(head);
1605         spin_unlock(&sysctl_lock);
1606 }
1607
1608 static struct ctl_table_set *
1609 lookup_header_set(struct ctl_table_root *root, struct nsproxy *namespaces)
1610 {
1611         struct ctl_table_set *set = &root->default_set;
1612         if (root->lookup)
1613                 set = root->lookup(root, namespaces);
1614         return set;
1615 }
1616
1617 static struct list_head *
1618 lookup_header_list(struct ctl_table_root *root, struct nsproxy *namespaces)
1619 {
1620         struct ctl_table_set *set = lookup_header_set(root, namespaces);
1621         return &set->list;
1622 }
1623
1624 struct ctl_table_header *__sysctl_head_next(struct nsproxy *namespaces,
1625                                             struct ctl_table_header *prev)
1626 {
1627         struct ctl_table_root *root;
1628         struct list_head *header_list;
1629         struct ctl_table_header *head;
1630         struct list_head *tmp;
1631
1632         spin_lock(&sysctl_lock);
1633         if (prev) {
1634                 head = prev;
1635                 tmp = &prev->ctl_entry;
1636                 unuse_table(prev);
1637                 goto next;
1638         }
1639         tmp = &root_table_header.ctl_entry;
1640         for (;;) {
1641                 head = list_entry(tmp, struct ctl_table_header, ctl_entry);
1642
1643                 if (!use_table(head))
1644                         goto next;
1645                 spin_unlock(&sysctl_lock);
1646                 return head;
1647         next:
1648                 root = head->root;
1649                 tmp = tmp->next;
1650                 header_list = lookup_header_list(root, namespaces);
1651                 if (tmp != header_list)
1652                         continue;
1653
1654                 do {
1655                         root = list_entry(root->root_list.next,
1656                                         struct ctl_table_root, root_list);
1657                         if (root == &sysctl_table_root)
1658                                 goto out;
1659                         header_list = lookup_header_list(root, namespaces);
1660                 } while (list_empty(header_list));
1661                 tmp = header_list->next;
1662         }
1663 out:
1664         spin_unlock(&sysctl_lock);
1665         return NULL;
1666 }
1667
1668 struct ctl_table_header *sysctl_head_next(struct ctl_table_header *prev)
1669 {
1670         return __sysctl_head_next(current->nsproxy, prev);
1671 }
1672
1673 void register_sysctl_root(struct ctl_table_root *root)
1674 {
1675         spin_lock(&sysctl_lock);
1676         list_add_tail(&root->root_list, &sysctl_table_root.root_list);
1677         spin_unlock(&sysctl_lock);
1678 }
1679
1680 /*
1681  * sysctl_perm does NOT grant the superuser all rights automatically, because
1682  * some sysctl variables are readonly even to root.
1683  */
1684
1685 static int test_perm(int mode, int op)
1686 {
1687         if (!current_euid())
1688                 mode >>= 6;
1689         else if (in_egroup_p(0))
1690                 mode >>= 3;
1691         if ((op & ~mode & (MAY_READ|MAY_WRITE|MAY_EXEC)) == 0)
1692                 return 0;
1693         return -EACCES;
1694 }
1695
1696 int sysctl_perm(struct ctl_table_root *root, struct ctl_table *table, int op)
1697 {
1698         int error;
1699         int mode;
1700
1701         error = security_sysctl(table, op & (MAY_READ | MAY_WRITE | MAY_EXEC));
1702         if (error)
1703                 return error;
1704
1705         if (root->permissions)
1706                 mode = root->permissions(root, current->nsproxy, table);
1707         else
1708                 mode = table->mode;
1709
1710         return test_perm(mode, op);
1711 }
1712
1713 static void sysctl_set_parent(struct ctl_table *parent, struct ctl_table *table)
1714 {
1715         for (; table->procname; table++) {
1716                 table->parent = parent;
1717                 if (table->child)
1718                         sysctl_set_parent(table, table->child);
1719         }
1720 }
1721
1722 static __init int sysctl_init(void)
1723 {
1724         sysctl_set_parent(NULL, root_table);
1725 #ifdef CONFIG_SYSCTL_SYSCALL_CHECK
1726         sysctl_check_table(current->nsproxy, root_table);
1727 #endif
1728         return 0;
1729 }
1730
1731 core_initcall(sysctl_init);
1732
1733 static struct ctl_table *is_branch_in(struct ctl_table *branch,
1734                                       struct ctl_table *table)
1735 {
1736         struct ctl_table *p;
1737         const char *s = branch->procname;
1738
1739         /* branch should have named subdirectory as its first element */
1740         if (!s || !branch->child)
1741                 return NULL;
1742
1743         /* ... and nothing else */
1744         if (branch[1].procname)
1745                 return NULL;
1746
1747         /* table should contain subdirectory with the same name */
1748         for (p = table; p->procname; p++) {
1749                 if (!p->child)
1750                         continue;
1751                 if (p->procname && strcmp(p->procname, s) == 0)
1752                         return p;
1753         }
1754         return NULL;
1755 }
1756
1757 /* see if attaching q to p would be an improvement */
1758 static void try_attach(struct ctl_table_header *p, struct ctl_table_header *q)
1759 {
1760         struct ctl_table *to = p->ctl_table, *by = q->ctl_table;
1761         struct ctl_table *next;
1762         int is_better = 0;
1763         int not_in_parent = !p->attached_by;
1764
1765         while ((next = is_branch_in(by, to)) != NULL) {
1766                 if (by == q->attached_by)
1767                         is_better = 1;
1768                 if (to == p->attached_by)
1769                         not_in_parent = 1;
1770                 by = by->child;
1771                 to = next->child;
1772         }
1773
1774         if (is_better && not_in_parent) {
1775                 q->attached_by = by;
1776                 q->attached_to = to;
1777                 q->parent = p;
1778         }
1779 }
1780
1781 /**
1782  * __register_sysctl_paths - register a sysctl hierarchy
1783  * @root: List of sysctl headers to register on
1784  * @namespaces: Data to compute which lists of sysctl entries are visible
1785  * @path: The path to the directory the sysctl table is in.
1786  * @table: the top-level table structure
1787  *
1788  * Register a sysctl table hierarchy. @table should be a filled in ctl_table
1789  * array. A completely 0 filled entry terminates the table.
1790  *
1791  * The members of the &struct ctl_table structure are used as follows:
1792  *
1793  * procname - the name of the sysctl file under /proc/sys. Set to %NULL to not
1794  *            enter a sysctl file
1795  *
1796  * data - a pointer to data for use by proc_handler
1797  *
1798  * maxlen - the maximum size in bytes of the data
1799  *
1800  * mode - the file permissions for the /proc/sys file, and for sysctl(2)
1801  *
1802  * child - a pointer to the child sysctl table if this entry is a directory, or
1803  *         %NULL.
1804  *
1805  * proc_handler - the text handler routine (described below)
1806  *
1807  * de - for internal use by the sysctl routines
1808  *
1809  * extra1, extra2 - extra pointers usable by the proc handler routines
1810  *
1811  * Leaf nodes in the sysctl tree will be represented by a single file
1812  * under /proc; non-leaf nodes will be represented by directories.
1813  *
1814  * sysctl(2) can automatically manage read and write requests through
1815  * the sysctl table.  The data and maxlen fields of the ctl_table
1816  * struct enable minimal validation of the values being written to be
1817  * performed, and the mode field allows minimal authentication.
1818  *
1819  * There must be a proc_handler routine for any terminal nodes
1820  * mirrored under /proc/sys (non-terminals are handled by a built-in
1821  * directory handler).  Several default handlers are available to
1822  * cover common cases -
1823  *
1824  * proc_dostring(), proc_dointvec(), proc_dointvec_jiffies(),
1825  * proc_dointvec_userhz_jiffies(), proc_dointvec_minmax(),
1826  * proc_doulongvec_ms_jiffies_minmax(), proc_doulongvec_minmax()
1827  *
1828  * It is the handler's job to read the input buffer from user memory
1829  * and process it. The handler should return 0 on success.
1830  *
1831  * This routine returns %NULL on a failure to register, and a pointer
1832  * to the table header on success.
1833  */
1834 struct ctl_table_header *__register_sysctl_paths(
1835         struct ctl_table_root *root,
1836         struct nsproxy *namespaces,
1837         const struct ctl_path *path, struct ctl_table *table)
1838 {
1839         struct ctl_table_header *header;
1840         struct ctl_table *new, **prevp;
1841         unsigned int n, npath;
1842         struct ctl_table_set *set;
1843
1844         /* Count the path components */
1845         for (npath = 0; path[npath].procname; ++npath)
1846                 ;
1847
1848         /*
1849          * For each path component, allocate a 2-element ctl_table array.
1850          * The first array element will be filled with the sysctl entry
1851          * for this, the second will be the sentinel (procname == 0).
1852          *
1853          * We allocate everything in one go so that we don't have to
1854          * worry about freeing additional memory in unregister_sysctl_table.
1855          */
1856         header = kzalloc(sizeof(struct ctl_table_header) +
1857                          (2 * npath * sizeof(struct ctl_table)), GFP_KERNEL);
1858         if (!header)
1859                 return NULL;
1860
1861         new = (struct ctl_table *) (header + 1);
1862
1863         /* Now connect the dots */
1864         prevp = &header->ctl_table;
1865         for (n = 0; n < npath; ++n, ++path) {
1866                 /* Copy the procname */
1867                 new->procname = path->procname;
1868                 new->mode     = 0555;
1869
1870                 *prevp = new;
1871                 prevp = &new->child;
1872
1873                 new += 2;
1874         }
1875         *prevp = table;
1876         header->ctl_table_arg = table;
1877
1878         INIT_LIST_HEAD(&header->ctl_entry);
1879         header->used = 0;
1880         header->unregistering = NULL;
1881         header->root = root;
1882         sysctl_set_parent(NULL, header->ctl_table);
1883         header->count = 1;
1884 #ifdef CONFIG_SYSCTL_SYSCALL_CHECK
1885         if (sysctl_check_table(namespaces, header->ctl_table)) {
1886                 kfree(header);
1887                 return NULL;
1888         }
1889 #endif
1890         spin_lock(&sysctl_lock);
1891         header->set = lookup_header_set(root, namespaces);
1892         header->attached_by = header->ctl_table;
1893         header->attached_to = root_table;
1894         header->parent = &root_table_header;
1895         for (set = header->set; set; set = set->parent) {
1896                 struct ctl_table_header *p;
1897                 list_for_each_entry(p, &set->list, ctl_entry) {
1898                         if (p->unregistering)
1899                                 continue;
1900                         try_attach(p, header);
1901                 }
1902         }
1903         header->parent->count++;
1904         list_add_tail(&header->ctl_entry, &header->set->list);
1905         spin_unlock(&sysctl_lock);
1906
1907         return header;
1908 }
1909
1910 /**
1911  * register_sysctl_table_path - register a sysctl table hierarchy
1912  * @path: The path to the directory the sysctl table is in.
1913  * @table: the top-level table structure
1914  *
1915  * Register a sysctl table hierarchy. @table should be a filled in ctl_table
1916  * array. A completely 0 filled entry terminates the table.
1917  *
1918  * See __register_sysctl_paths for more details.
1919  */
1920 struct ctl_table_header *register_sysctl_paths(const struct ctl_path *path,
1921                                                 struct ctl_table *table)
1922 {
1923         return __register_sysctl_paths(&sysctl_table_root, current->nsproxy,
1924                                         path, table);
1925 }
1926
1927 /**
1928  * register_sysctl_table - register a sysctl table hierarchy
1929  * @table: the top-level table structure
1930  *
1931  * Register a sysctl table hierarchy. @table should be a filled in ctl_table
1932  * array. A completely 0 filled entry terminates the table.
1933  *
1934  * See register_sysctl_paths for more details.
1935  */
1936 struct ctl_table_header *register_sysctl_table(struct ctl_table *table)
1937 {
1938         static const struct ctl_path null_path[] = { {} };
1939
1940         return register_sysctl_paths(null_path, table);
1941 }
1942
1943 /**
1944  * unregister_sysctl_table - unregister a sysctl table hierarchy
1945  * @header: the header returned from register_sysctl_table
1946  *
1947  * Unregisters the sysctl table and all children. proc entries may not
1948  * actually be removed until they are no longer used by anyone.
1949  */
1950 void unregister_sysctl_table(struct ctl_table_header * header)
1951 {
1952         might_sleep();
1953
1954         if (header == NULL)
1955                 return;
1956
1957         spin_lock(&sysctl_lock);
1958         start_unregistering(header);
1959         if (!--header->parent->count) {
1960                 WARN_ON(1);
1961                 call_rcu(&header->parent->rcu, free_head);
1962         }
1963         if (!--header->count)
1964                 call_rcu(&header->rcu, free_head);
1965         spin_unlock(&sysctl_lock);
1966 }
1967
1968 int sysctl_is_seen(struct ctl_table_header *p)
1969 {
1970         struct ctl_table_set *set = p->set;
1971         int res;
1972         spin_lock(&sysctl_lock);
1973         if (p->unregistering)
1974                 res = 0;
1975         else if (!set->is_seen)
1976                 res = 1;
1977         else
1978                 res = set->is_seen(set);
1979         spin_unlock(&sysctl_lock);
1980         return res;
1981 }
1982
1983 void setup_sysctl_set(struct ctl_table_set *p,
1984         struct ctl_table_set *parent,
1985         int (*is_seen)(struct ctl_table_set *))
1986 {
1987         INIT_LIST_HEAD(&p->list);
1988         p->parent = parent ? parent : &sysctl_table_root.default_set;
1989         p->is_seen = is_seen;
1990 }
1991
1992 #else /* !CONFIG_SYSCTL */
1993 struct ctl_table_header *register_sysctl_table(struct ctl_table * table)
1994 {
1995         return NULL;
1996 }
1997
1998 struct ctl_table_header *register_sysctl_paths(const struct ctl_path *path,
1999                                                     struct ctl_table *table)
2000 {
2001         return NULL;
2002 }
2003
2004 void unregister_sysctl_table(struct ctl_table_header * table)
2005 {
2006 }
2007
2008 void setup_sysctl_set(struct ctl_table_set *p,
2009         struct ctl_table_set *parent,
2010         int (*is_seen)(struct ctl_table_set *))
2011 {
2012 }
2013
2014 void sysctl_head_put(struct ctl_table_header *head)
2015 {
2016 }
2017
2018 #endif /* CONFIG_SYSCTL */
2019
2020 /*
2021  * /proc/sys support
2022  */
2023
2024 #ifdef CONFIG_PROC_SYSCTL
2025
2026 static int _proc_do_string(void* data, int maxlen, int write,
2027                            void __user *buffer,
2028                            size_t *lenp, loff_t *ppos)
2029 {
2030         size_t len;
2031         char __user *p;
2032         char c;
2033
2034         if (!data || !maxlen || !*lenp) {
2035                 *lenp = 0;
2036                 return 0;
2037         }
2038
2039         if (write) {
2040                 len = 0;
2041                 p = buffer;
2042                 while (len < *lenp) {
2043                         if (get_user(c, p++))
2044                                 return -EFAULT;
2045                         if (c == 0 || c == '\n')
2046                                 break;
2047                         len++;
2048                 }
2049                 if (len >= maxlen)
2050                         len = maxlen-1;
2051                 if(copy_from_user(data, buffer, len))
2052                         return -EFAULT;
2053                 ((char *) data)[len] = 0;
2054                 *ppos += *lenp;
2055         } else {
2056                 len = strlen(data);
2057                 if (len > maxlen)
2058                         len = maxlen;
2059
2060                 if (*ppos > len) {
2061                         *lenp = 0;
2062                         return 0;
2063                 }
2064
2065                 data += *ppos;
2066                 len  -= *ppos;
2067
2068                 if (len > *lenp)
2069                         len = *lenp;
2070                 if (len)
2071                         if(copy_to_user(buffer, data, len))
2072                                 return -EFAULT;
2073                 if (len < *lenp) {
2074                         if(put_user('\n', ((char __user *) buffer) + len))
2075                                 return -EFAULT;
2076                         len++;
2077                 }
2078                 *lenp = len;
2079                 *ppos += len;
2080         }
2081         return 0;
2082 }
2083
2084 /**
2085  * proc_dostring - read a string sysctl
2086  * @table: the sysctl table
2087  * @write: %TRUE if this is a write to the sysctl file
2088  * @buffer: the user buffer
2089  * @lenp: the size of the user buffer
2090  * @ppos: file position
2091  *
2092  * Reads/writes a string from/to the user buffer. If the kernel
2093  * buffer provided is not large enough to hold the string, the
2094  * string is truncated. The copied string is %NULL-terminated.
2095  * If the string is being read by the user process, it is copied
2096  * and a newline '\n' is added. It is truncated if the buffer is
2097  * not large enough.
2098  *
2099  * Returns 0 on success.
2100  */
2101 int proc_dostring(struct ctl_table *table, int write,
2102                   void __user *buffer, size_t *lenp, loff_t *ppos)
2103 {
2104         return _proc_do_string(table->data, table->maxlen, write,
2105                                buffer, lenp, ppos);
2106 }
2107
2108 static size_t proc_skip_spaces(char **buf)
2109 {
2110         size_t ret;
2111         char *tmp = skip_spaces(*buf);
2112         ret = tmp - *buf;
2113         *buf = tmp;
2114         return ret;
2115 }
2116
2117 static void proc_skip_char(char **buf, size_t *size, const char v)
2118 {
2119         while (*size) {
2120                 if (**buf != v)
2121                         break;
2122                 (*size)--;
2123                 (*buf)++;
2124         }
2125 }
2126
2127 #define TMPBUFLEN 22
2128 /**
2129  * proc_get_long - reads an ASCII formatted integer from a user buffer
2130  *
2131  * @buf: a kernel buffer
2132  * @size: size of the kernel buffer
2133  * @val: this is where the number will be stored
2134  * @neg: set to %TRUE if number is negative
2135  * @perm_tr: a vector which contains the allowed trailers
2136  * @perm_tr_len: size of the perm_tr vector
2137  * @tr: pointer to store the trailer character
2138  *
2139  * In case of success %0 is returned and @buf and @size are updated with
2140  * the amount of bytes read. If @tr is non-NULL and a trailing
2141  * character exists (size is non-zero after returning from this
2142  * function), @tr is updated with the trailing character.
2143  */
2144 static int proc_get_long(char **buf, size_t *size,
2145                           unsigned long *val, bool *neg,
2146                           const char *perm_tr, unsigned perm_tr_len, char *tr)
2147 {
2148         int len;
2149         char *p, tmp[TMPBUFLEN];
2150
2151         if (!*size)
2152                 return -EINVAL;
2153
2154         len = *size;
2155         if (len > TMPBUFLEN - 1)
2156                 len = TMPBUFLEN - 1;
2157
2158         memcpy(tmp, *buf, len);
2159
2160         tmp[len] = 0;
2161         p = tmp;
2162         if (*p == '-' && *size > 1) {
2163                 *neg = true;
2164                 p++;
2165         } else
2166                 *neg = false;
2167         if (!isdigit(*p))
2168                 return -EINVAL;
2169
2170         *val = simple_strtoul(p, &p, 0);
2171
2172         len = p - tmp;
2173
2174         /* We don't know if the next char is whitespace thus we may accept
2175          * invalid integers (e.g. 1234...a) or two integers instead of one
2176          * (e.g. 123...1). So lets not allow such large numbers. */
2177         if (len == TMPBUFLEN - 1)
2178                 return -EINVAL;
2179
2180         if (len < *size && perm_tr_len && !memchr(perm_tr, *p, perm_tr_len))
2181                 return -EINVAL;
2182
2183         if (tr && (len < *size))
2184                 *tr = *p;
2185
2186         *buf += len;
2187         *size -= len;
2188
2189         return 0;
2190 }
2191
2192 /**
2193  * proc_put_long - converts an integer to a decimal ASCII formatted string
2194  *
2195  * @buf: the user buffer
2196  * @size: the size of the user buffer
2197  * @val: the integer to be converted
2198  * @neg: sign of the number, %TRUE for negative
2199  *
2200  * In case of success %0 is returned and @buf and @size are updated with
2201  * the amount of bytes written.
2202  */
2203 static int proc_put_long(void __user **buf, size_t *size, unsigned long val,
2204                           bool neg)
2205 {
2206         int len;
2207         char tmp[TMPBUFLEN], *p = tmp;
2208
2209         sprintf(p, "%s%lu", neg ? "-" : "", val);
2210         len = strlen(tmp);
2211         if (len > *size)
2212                 len = *size;
2213         if (copy_to_user(*buf, tmp, len))
2214                 return -EFAULT;
2215         *size -= len;
2216         *buf += len;
2217         return 0;
2218 }
2219 #undef TMPBUFLEN
2220
2221 static int proc_put_char(void __user **buf, size_t *size, char c)
2222 {
2223         if (*size) {
2224                 char __user **buffer = (char __user **)buf;
2225                 if (put_user(c, *buffer))
2226                         return -EFAULT;
2227                 (*size)--, (*buffer)++;
2228                 *buf = *buffer;
2229         }
2230         return 0;
2231 }
2232
2233 static int do_proc_dointvec_conv(bool *negp, unsigned long *lvalp,
2234                                  int *valp,
2235                                  int write, void *data)
2236 {
2237         if (write) {
2238                 *valp = *negp ? -*lvalp : *lvalp;
2239         } else {
2240                 int val = *valp;
2241                 if (val < 0) {
2242                         *negp = true;
2243                         *lvalp = (unsigned long)-val;
2244                 } else {
2245                         *negp = false;
2246                         *lvalp = (unsigned long)val;
2247                 }
2248         }
2249         return 0;
2250 }
2251
2252 static const char proc_wspace_sep[] = { ' ', '\t', '\n' };
2253
2254 static int __do_proc_dointvec(void *tbl_data, struct ctl_table *table,
2255                   int write, void __user *buffer,
2256                   size_t *lenp, loff_t *ppos,
2257                   int (*conv)(bool *negp, unsigned long *lvalp, int *valp,
2258                               int write, void *data),
2259                   void *data)
2260 {
2261         int *i, vleft, first = 1, err = 0;
2262         unsigned long page = 0;
2263         size_t left;
2264         char *kbuf;
2265
2266         if (!tbl_data || !table->maxlen || !*lenp || (*ppos && !write)) {
2267                 *lenp = 0;
2268                 return 0;
2269         }
2270
2271         i = (int *) tbl_data;
2272         vleft = table->maxlen / sizeof(*i);
2273         left = *lenp;
2274
2275         if (!conv)
2276                 conv = do_proc_dointvec_conv;
2277
2278         if (write) {
2279                 if (left > PAGE_SIZE - 1)
2280                         left = PAGE_SIZE - 1;
2281                 page = __get_free_page(GFP_TEMPORARY);
2282                 kbuf = (char *) page;
2283                 if (!kbuf)
2284                         return -ENOMEM;
2285                 if (copy_from_user(kbuf, buffer, left)) {
2286                         err = -EFAULT;
2287                         goto free;
2288                 }
2289                 kbuf[left] = 0;
2290         }
2291
2292         for (; left && vleft--; i++, first=0) {
2293                 unsigned long lval;
2294                 bool neg;
2295
2296                 if (write) {
2297                         left -= proc_skip_spaces(&kbuf);
2298
2299                         if (!left)
2300                                 break;
2301                         err = proc_get_long(&kbuf, &left, &lval, &neg,
2302                                              proc_wspace_sep,
2303                                              sizeof(proc_wspace_sep), NULL);
2304                         if (err)
2305                                 break;
2306                         if (conv(&neg, &lval, i, 1, data)) {
2307                                 err = -EINVAL;
2308                                 break;
2309                         }
2310                 } else {
2311                         if (conv(&neg, &lval, i, 0, data)) {
2312                                 err = -EINVAL;
2313                                 break;
2314                         }
2315                         if (!first)
2316                                 err = proc_put_char(&buffer, &left, '\t');
2317                         if (err)
2318                                 break;
2319                         err = proc_put_long(&buffer, &left, lval, neg);
2320                         if (err)
2321                                 break;
2322                 }
2323         }
2324
2325         if (!write && !first && left && !err)
2326                 err = proc_put_char(&buffer, &left, '\n');
2327         if (write && !err && left)
2328                 left -= proc_skip_spaces(&kbuf);
2329 free:
2330         if (write) {
2331                 free_page(page);
2332                 if (first)
2333                         return err ? : -EINVAL;
2334         }
2335         *lenp -= left;
2336         *ppos += *lenp;
2337         return err;
2338 }
2339
2340 static int do_proc_dointvec(struct ctl_table *table, int write,
2341                   void __user *buffer, size_t *lenp, loff_t *ppos,
2342                   int (*conv)(bool *negp, unsigned long *lvalp, int *valp,
2343                               int write, void *data),
2344                   void *data)
2345 {
2346         return __do_proc_dointvec(table->data, table, write,
2347                         buffer, lenp, ppos, conv, data);
2348 }
2349
2350 /**
2351  * proc_dointvec - read a vector of integers
2352  * @table: the sysctl table
2353  * @write: %TRUE if this is a write to the sysctl file
2354  * @buffer: the user buffer
2355  * @lenp: the size of the user buffer
2356  * @ppos: file position
2357  *
2358  * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2359  * values from/to the user buffer, treated as an ASCII string.
2360  *
2361  * Returns 0 on success.
2362  */
2363 int proc_dointvec(struct ctl_table *table, int write,
2364                      void __user *buffer, size_t *lenp, loff_t *ppos)
2365 {
2366     return do_proc_dointvec(table,write,buffer,lenp,ppos,
2367                             NULL,NULL);
2368 }
2369
2370 /*
2371  * Taint values can only be increased
2372  * This means we can safely use a temporary.
2373  */
2374 static int proc_taint(struct ctl_table *table, int write,
2375                                void __user *buffer, size_t *lenp, loff_t *ppos)
2376 {
2377         struct ctl_table t;
2378         unsigned long tmptaint = get_taint();
2379         int err;
2380
2381         if (write && !capable(CAP_SYS_ADMIN))
2382                 return -EPERM;
2383
2384         t = *table;
2385         t.data = &tmptaint;
2386         err = proc_doulongvec_minmax(&t, write, buffer, lenp, ppos);
2387         if (err < 0)
2388                 return err;
2389
2390         if (write) {
2391                 /*
2392                  * Poor man's atomic or. Not worth adding a primitive
2393                  * to everyone's atomic.h for this
2394                  */
2395                 int i;
2396                 for (i = 0; i < BITS_PER_LONG && tmptaint >> i; i++) {
2397                         if ((tmptaint >> i) & 1)
2398                                 add_taint(i);
2399                 }
2400         }
2401
2402         return err;
2403 }
2404
2405 #ifdef CONFIG_PRINTK
2406 static int proc_dmesg_restrict(struct ctl_table *table, int write,
2407                                 void __user *buffer, size_t *lenp, loff_t *ppos)
2408 {
2409         if (write && !capable(CAP_SYS_ADMIN))
2410                 return -EPERM;
2411
2412         return proc_dointvec_minmax(table, write, buffer, lenp, ppos);
2413 }
2414 #endif
2415
2416 struct do_proc_dointvec_minmax_conv_param {
2417         int *min;
2418         int *max;
2419 };
2420
2421 static int do_proc_dointvec_minmax_conv(bool *negp, unsigned long *lvalp,
2422                                         int *valp,
2423                                         int write, void *data)
2424 {
2425         struct do_proc_dointvec_minmax_conv_param *param = data;
2426         if (write) {
2427                 int val = *negp ? -*lvalp : *lvalp;
2428                 if ((param->min && *param->min > val) ||
2429                     (param->max && *param->max < val))
2430                         return -EINVAL;
2431                 *valp = val;
2432         } else {
2433                 int val = *valp;
2434                 if (val < 0) {
2435                         *negp = true;
2436                         *lvalp = (unsigned long)-val;
2437                 } else {
2438                         *negp = false;
2439                         *lvalp = (unsigned long)val;
2440                 }
2441         }
2442         return 0;
2443 }
2444
2445 /**
2446  * proc_dointvec_minmax - read a vector of integers with min/max values
2447  * @table: the sysctl table
2448  * @write: %TRUE if this is a write to the sysctl file
2449  * @buffer: the user buffer
2450  * @lenp: the size of the user buffer
2451  * @ppos: file position
2452  *
2453  * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2454  * values from/to the user buffer, treated as an ASCII string.
2455  *
2456  * This routine will ensure the values are within the range specified by
2457  * table->extra1 (min) and table->extra2 (max).
2458  *
2459  * Returns 0 on success.
2460  */
2461 int proc_dointvec_minmax(struct ctl_table *table, int write,
2462                   void __user *buffer, size_t *lenp, loff_t *ppos)
2463 {
2464         struct do_proc_dointvec_minmax_conv_param param = {
2465                 .min = (int *) table->extra1,
2466                 .max = (int *) table->extra2,
2467         };
2468         return do_proc_dointvec(table, write, buffer, lenp, ppos,
2469                                 do_proc_dointvec_minmax_conv, &param);
2470 }
2471
2472 static int __do_proc_doulongvec_minmax(void *data, struct ctl_table *table, int write,
2473                                      void __user *buffer,
2474                                      size_t *lenp, loff_t *ppos,
2475                                      unsigned long convmul,
2476                                      unsigned long convdiv)
2477 {
2478         unsigned long *i, *min, *max;
2479         int vleft, first = 1, err = 0;
2480         unsigned long page = 0;
2481         size_t left;
2482         char *kbuf;
2483
2484         if (!data || !table->maxlen || !*lenp || (*ppos && !write)) {
2485                 *lenp = 0;
2486                 return 0;
2487         }
2488
2489         i = (unsigned long *) data;
2490         min = (unsigned long *) table->extra1;
2491         max = (unsigned long *) table->extra2;
2492         vleft = table->maxlen / sizeof(unsigned long);
2493         left = *lenp;
2494
2495         if (write) {
2496                 if (left > PAGE_SIZE - 1)
2497                         left = PAGE_SIZE - 1;
2498                 page = __get_free_page(GFP_TEMPORARY);
2499                 kbuf = (char *) page;
2500                 if (!kbuf)
2501                         return -ENOMEM;
2502                 if (copy_from_user(kbuf, buffer, left)) {
2503                         err = -EFAULT;
2504                         goto free;
2505                 }
2506                 kbuf[left] = 0;
2507         }
2508
2509         for (; left && vleft--; i++, first = 0) {
2510                 unsigned long val;
2511
2512                 if (write) {
2513                         bool neg;
2514
2515                         left -= proc_skip_spaces(&kbuf);
2516
2517                         err = proc_get_long(&kbuf, &left, &val, &neg,
2518                                              proc_wspace_sep,
2519                                              sizeof(proc_wspace_sep), NULL);
2520                         if (err)
2521                                 break;
2522                         if (neg)
2523                                 continue;
2524                         if ((min && val < *min) || (max && val > *max))
2525                                 continue;
2526                         *i = val;
2527                 } else {
2528                         val = convdiv * (*i) / convmul;
2529                         if (!first)
2530                                 err = proc_put_char(&buffer, &left, '\t');
2531                         err = proc_put_long(&buffer, &left, val, false);
2532                         if (err)
2533                                 break;
2534                 }
2535         }
2536
2537         if (!write && !first && left && !err)
2538                 err = proc_put_char(&buffer, &left, '\n');
2539         if (write && !err)
2540                 left -= proc_skip_spaces(&kbuf);
2541 free:
2542         if (write) {
2543                 free_page(page);
2544                 if (first)
2545                         return err ? : -EINVAL;
2546         }
2547         *lenp -= left;
2548         *ppos += *lenp;
2549         return err;
2550 }
2551
2552 static int do_proc_doulongvec_minmax(struct ctl_table *table, int write,
2553                                      void __user *buffer,
2554                                      size_t *lenp, loff_t *ppos,
2555                                      unsigned long convmul,
2556                                      unsigned long convdiv)
2557 {
2558         return __do_proc_doulongvec_minmax(table->data, table, write,
2559                         buffer, lenp, ppos, convmul, convdiv);
2560 }
2561
2562 /**
2563  * proc_doulongvec_minmax - read a vector of long integers with min/max values
2564  * @table: the sysctl table
2565  * @write: %TRUE if this is a write to the sysctl file
2566  * @buffer: the user buffer
2567  * @lenp: the size of the user buffer
2568  * @ppos: file position
2569  *
2570  * Reads/writes up to table->maxlen/sizeof(unsigned long) unsigned long
2571  * values from/to the user buffer, treated as an ASCII string.
2572  *
2573  * This routine will ensure the values are within the range specified by
2574  * table->extra1 (min) and table->extra2 (max).
2575  *
2576  * Returns 0 on success.
2577  */
2578 int proc_doulongvec_minmax(struct ctl_table *table, int write,
2579                            void __user *buffer, size_t *lenp, loff_t *ppos)
2580 {
2581     return do_proc_doulongvec_minmax(table, write, buffer, lenp, ppos, 1l, 1l);
2582 }
2583
2584 /**
2585  * proc_doulongvec_ms_jiffies_minmax - read a vector of millisecond values with min/max values
2586  * @table: the sysctl table
2587  * @write: %TRUE if this is a write to the sysctl file
2588  * @buffer: the user buffer
2589  * @lenp: the size of the user buffer
2590  * @ppos: file position
2591  *
2592  * Reads/writes up to table->maxlen/sizeof(unsigned long) unsigned long
2593  * values from/to the user buffer, treated as an ASCII string. The values
2594  * are treated as milliseconds, and converted to jiffies when they are stored.
2595  *
2596  * This routine will ensure the values are within the range specified by
2597  * table->extra1 (min) and table->extra2 (max).
2598  *
2599  * Returns 0 on success.
2600  */
2601 int proc_doulongvec_ms_jiffies_minmax(struct ctl_table *table, int write,
2602                                       void __user *buffer,
2603                                       size_t *lenp, loff_t *ppos)
2604 {
2605     return do_proc_doulongvec_minmax(table, write, buffer,
2606                                      lenp, ppos, HZ, 1000l);
2607 }
2608
2609
2610 static int do_proc_dointvec_jiffies_conv(bool *negp, unsigned long *lvalp,
2611                                          int *valp,
2612                                          int write, void *data)
2613 {
2614         if (write) {
2615                 if (*lvalp > LONG_MAX / HZ)
2616                         return 1;
2617                 *valp = *negp ? -(*lvalp*HZ) : (*lvalp*HZ);
2618         } else {
2619                 int val = *valp;
2620                 unsigned long lval;
2621                 if (val < 0) {
2622                         *negp = true;
2623                         lval = (unsigned long)-val;
2624                 } else {
2625                         *negp = false;
2626                         lval = (unsigned long)val;
2627                 }
2628                 *lvalp = lval / HZ;
2629         }
2630         return 0;
2631 }
2632
2633 static int do_proc_dointvec_userhz_jiffies_conv(bool *negp, unsigned long *lvalp,
2634                                                 int *valp,
2635                                                 int write, void *data)
2636 {
2637         if (write) {
2638                 if (USER_HZ < HZ && *lvalp > (LONG_MAX / HZ) * USER_HZ)
2639                         return 1;
2640                 *valp = clock_t_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_clock_t(lval);
2652         }
2653         return 0;
2654 }
2655
2656 static int do_proc_dointvec_ms_jiffies_conv(bool *negp, unsigned long *lvalp,
2657                                             int *valp,
2658                                             int write, void *data)
2659 {
2660         if (write) {
2661                 *valp = msecs_to_jiffies(*negp ? -*lvalp : *lvalp);
2662         } else {
2663                 int val = *valp;
2664                 unsigned long lval;
2665                 if (val < 0) {
2666                         *negp = true;
2667                         lval = (unsigned long)-val;
2668                 } else {
2669                         *negp = false;
2670                         lval = (unsigned long)val;
2671                 }
2672                 *lvalp = jiffies_to_msecs(lval);
2673         }
2674         return 0;
2675 }
2676
2677 /**
2678  * proc_dointvec_jiffies - read a vector of integers as seconds
2679  * @table: the sysctl table
2680  * @write: %TRUE if this is a write to the sysctl file
2681  * @buffer: the user buffer
2682  * @lenp: the size of the user buffer
2683  * @ppos: file position
2684  *
2685  * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2686  * values from/to the user buffer, treated as an ASCII string.
2687  * The values read are assumed to be in seconds, and are converted into
2688  * jiffies.
2689  *
2690  * Returns 0 on success.
2691  */
2692 int proc_dointvec_jiffies(struct ctl_table *table, int write,
2693                           void __user *buffer, size_t *lenp, loff_t *ppos)
2694 {
2695     return do_proc_dointvec(table,write,buffer,lenp,ppos,
2696                             do_proc_dointvec_jiffies_conv,NULL);
2697 }
2698
2699 /**
2700  * proc_dointvec_userhz_jiffies - read a vector of integers as 1/USER_HZ seconds
2701  * @table: the sysctl table
2702  * @write: %TRUE if this is a write to the sysctl file
2703  * @buffer: the user buffer
2704  * @lenp: the size of the user buffer
2705  * @ppos: pointer to the file position
2706  *
2707  * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2708  * values from/to the user buffer, treated as an ASCII string.
2709  * The values read are assumed to be in 1/USER_HZ seconds, and
2710  * are converted into jiffies.
2711  *
2712  * Returns 0 on success.
2713  */
2714 int proc_dointvec_userhz_jiffies(struct ctl_table *table, int write,
2715                                  void __user *buffer, size_t *lenp, loff_t *ppos)
2716 {
2717     return do_proc_dointvec(table,write,buffer,lenp,ppos,
2718                             do_proc_dointvec_userhz_jiffies_conv,NULL);
2719 }
2720
2721 /**
2722  * proc_dointvec_ms_jiffies - read a vector of integers as 1 milliseconds
2723  * @table: the sysctl table
2724  * @write: %TRUE if this is a write to the sysctl file
2725  * @buffer: the user buffer
2726  * @lenp: the size of the user buffer
2727  * @ppos: file position
2728  * @ppos: the current position in the file
2729  *
2730  * Reads/writes up to table->maxlen/sizeof(unsigned int) integer
2731  * values from/to the user buffer, treated as an ASCII string.
2732  * The values read are assumed to be in 1/1000 seconds, and
2733  * are converted into jiffies.
2734  *
2735  * Returns 0 on success.
2736  */
2737 int proc_dointvec_ms_jiffies(struct ctl_table *table, int write,
2738                              void __user *buffer, size_t *lenp, loff_t *ppos)
2739 {
2740         return do_proc_dointvec(table, write, buffer, lenp, ppos,
2741                                 do_proc_dointvec_ms_jiffies_conv, NULL);
2742 }
2743
2744 static int proc_do_cad_pid(struct ctl_table *table, int write,
2745                            void __user *buffer, size_t *lenp, loff_t *ppos)
2746 {
2747         struct pid *new_pid;
2748         pid_t tmp;
2749         int r;
2750
2751         tmp = pid_vnr(cad_pid);
2752
2753         r = __do_proc_dointvec(&tmp, table, write, buffer,
2754                                lenp, ppos, NULL, NULL);
2755         if (r || !write)
2756                 return r;
2757
2758         new_pid = find_get_pid(tmp);
2759         if (!new_pid)
2760                 return -ESRCH;
2761
2762         put_pid(xchg(&cad_pid, new_pid));
2763         return 0;
2764 }
2765
2766 /**
2767  * proc_do_large_bitmap - read/write from/to a large bitmap
2768  * @table: the sysctl table
2769  * @write: %TRUE if this is a write to the sysctl file
2770  * @buffer: the user buffer
2771  * @lenp: the size of the user buffer
2772  * @ppos: file position
2773  *
2774  * The bitmap is stored at table->data and the bitmap length (in bits)
2775  * in table->maxlen.
2776  *
2777  * We use a range comma separated format (e.g. 1,3-4,10-10) so that
2778  * large bitmaps may be represented in a compact manner. Writing into
2779  * the file will clear the bitmap then update it with the given input.
2780  *
2781  * Returns 0 on success.
2782  */
2783 int proc_do_large_bitmap(struct ctl_table *table, int write,
2784                          void __user *buffer, size_t *lenp, loff_t *ppos)
2785 {
2786         int err = 0;
2787         bool first = 1;
2788         size_t left = *lenp;
2789         unsigned long bitmap_len = table->maxlen;
2790         unsigned long *bitmap = (unsigned long *) table->data;
2791         unsigned long *tmp_bitmap = NULL;
2792         char tr_a[] = { '-', ',', '\n' }, tr_b[] = { ',', '\n', 0 }, c;
2793
2794         if (!bitmap_len || !left || (*ppos && !write)) {
2795                 *lenp = 0;
2796                 return 0;
2797         }
2798
2799         if (write) {
2800                 unsigned long page = 0;
2801                 char *kbuf;
2802
2803                 if (left > PAGE_SIZE - 1)
2804                         left = PAGE_SIZE - 1;
2805
2806                 page = __get_free_page(GFP_TEMPORARY);
2807                 kbuf = (char *) page;
2808                 if (!kbuf)
2809                         return -ENOMEM;
2810                 if (copy_from_user(kbuf, buffer, left)) {
2811                         free_page(page);
2812                         return -EFAULT;
2813                 }
2814                 kbuf[left] = 0;
2815
2816                 tmp_bitmap = kzalloc(BITS_TO_LONGS(bitmap_len) * sizeof(unsigned long),
2817                                      GFP_KERNEL);
2818                 if (!tmp_bitmap) {
2819                         free_page(page);
2820                         return -ENOMEM;
2821                 }
2822                 proc_skip_char(&kbuf, &left, '\n');
2823                 while (!err && left) {
2824                         unsigned long val_a, val_b;
2825                         bool neg;
2826
2827                         err = proc_get_long(&kbuf, &left, &val_a, &neg, tr_a,
2828                                              sizeof(tr_a), &c);
2829                         if (err)
2830                                 break;
2831                         if (val_a >= bitmap_len || neg) {
2832                                 err = -EINVAL;
2833                                 break;
2834                         }
2835
2836                         val_b = val_a;
2837                         if (left) {
2838                                 kbuf++;
2839                                 left--;
2840                         }
2841
2842                         if (c == '-') {
2843                                 err = proc_get_long(&kbuf, &left, &val_b,
2844                                                      &neg, tr_b, sizeof(tr_b),
2845                                                      &c);
2846                                 if (err)
2847                                         break;
2848                                 if (val_b >= bitmap_len || neg ||
2849                                     val_a > val_b) {
2850                                         err = -EINVAL;
2851                                         break;
2852                                 }
2853                                 if (left) {
2854                                         kbuf++;
2855                                         left--;
2856                                 }
2857                         }
2858
2859                         while (val_a <= val_b)
2860                                 set_bit(val_a++, tmp_bitmap);
2861
2862                         first = 0;
2863                         proc_skip_char(&kbuf, &left, '\n');
2864                 }
2865                 free_page(page);
2866         } else {
2867                 unsigned long bit_a, bit_b = 0;
2868
2869                 while (left) {
2870                         bit_a = find_next_bit(bitmap, bitmap_len, bit_b);
2871                         if (bit_a >= bitmap_len)
2872                                 break;
2873                         bit_b = find_next_zero_bit(bitmap, bitmap_len,
2874                                                    bit_a + 1) - 1;
2875
2876                         if (!first) {
2877                                 err = proc_put_char(&buffer, &left, ',');
2878                                 if (err)
2879                                         break;
2880                         }
2881                         err = proc_put_long(&buffer, &left, bit_a, false);
2882                         if (err)
2883                                 break;
2884                         if (bit_a != bit_b) {
2885                                 err = proc_put_char(&buffer, &left, '-');
2886                                 if (err)
2887                                         break;
2888                                 err = proc_put_long(&buffer, &left, bit_b, false);
2889                                 if (err)
2890                                         break;
2891                         }
2892
2893                         first = 0; bit_b++;
2894                 }
2895                 if (!err)
2896                         err = proc_put_char(&buffer, &left, '\n');
2897         }
2898
2899         if (!err) {
2900                 if (write) {
2901                         if (*ppos)
2902                                 bitmap_or(bitmap, bitmap, tmp_bitmap, bitmap_len);
2903                         else
2904                                 memcpy(bitmap, tmp_bitmap,
2905                                         BITS_TO_LONGS(bitmap_len) * sizeof(unsigned long));
2906                 }
2907                 kfree(tmp_bitmap);
2908                 *lenp -= left;
2909                 *ppos += *lenp;
2910                 return 0;
2911         } else {
2912                 kfree(tmp_bitmap);
2913                 return err;
2914         }
2915 }
2916
2917 #else /* CONFIG_PROC_SYSCTL */
2918
2919 int proc_dostring(struct ctl_table *table, int write,
2920                   void __user *buffer, size_t *lenp, loff_t *ppos)
2921 {
2922         return -ENOSYS;
2923 }
2924
2925 int proc_dointvec(struct ctl_table *table, int write,
2926                   void __user *buffer, size_t *lenp, loff_t *ppos)
2927 {
2928         return -ENOSYS;
2929 }
2930
2931 int proc_dointvec_minmax(struct ctl_table *table, int write,
2932                     void __user *buffer, size_t *lenp, loff_t *ppos)
2933 {
2934         return -ENOSYS;
2935 }
2936
2937 int proc_dointvec_jiffies(struct ctl_table *table, int write,
2938                     void __user *buffer, size_t *lenp, loff_t *ppos)
2939 {
2940         return -ENOSYS;
2941 }
2942
2943 int proc_dointvec_userhz_jiffies(struct ctl_table *table, int write,
2944                     void __user *buffer, size_t *lenp, loff_t *ppos)
2945 {
2946         return -ENOSYS;
2947 }
2948
2949 int proc_dointvec_ms_jiffies(struct ctl_table *table, int write,
2950                              void __user *buffer, size_t *lenp, loff_t *ppos)
2951 {
2952         return -ENOSYS;
2953 }
2954
2955 int proc_doulongvec_minmax(struct ctl_table *table, int write,
2956                     void __user *buffer, size_t *lenp, loff_t *ppos)
2957 {
2958         return -ENOSYS;
2959 }
2960
2961 int proc_doulongvec_ms_jiffies_minmax(struct ctl_table *table, int write,
2962                                       void __user *buffer,
2963                                       size_t *lenp, loff_t *ppos)
2964 {
2965     return -ENOSYS;
2966 }
2967
2968
2969 #endif /* CONFIG_PROC_SYSCTL */
2970
2971 /*
2972  * No sense putting this after each symbol definition, twice,
2973  * exception granted :-)
2974  */
2975 EXPORT_SYMBOL(proc_dointvec);
2976 EXPORT_SYMBOL(proc_dointvec_jiffies);
2977 EXPORT_SYMBOL(proc_dointvec_minmax);
2978 EXPORT_SYMBOL(proc_dointvec_userhz_jiffies);
2979 EXPORT_SYMBOL(proc_dointvec_ms_jiffies);
2980 EXPORT_SYMBOL(proc_dostring);
2981 EXPORT_SYMBOL(proc_doulongvec_minmax);
2982 EXPORT_SYMBOL(proc_doulongvec_ms_jiffies_minmax);
2983 EXPORT_SYMBOL(register_sysctl_table);
2984 EXPORT_SYMBOL(register_sysctl_paths);
2985 EXPORT_SYMBOL(unregister_sysctl_table);