linux/kernel/module.c
<<
>>
Prefs
   1/*
   2   Copyright (C) 2002 Richard Henderson
   3   Copyright (C) 2001 Rusty Russell, 2002, 2010 Rusty Russell IBM.
   4
   5    This program is free software; you can redistribute it and/or modify
   6    it under the terms of the GNU General Public License as published by
   7    the Free Software Foundation; either version 2 of the License, or
   8    (at your option) any later version.
   9
  10    This program is distributed in the hope that it will be useful,
  11    but WITHOUT ANY WARRANTY; without even the implied warranty of
  12    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  13    GNU General Public License for more details.
  14
  15    You should have received a copy of the GNU General Public License
  16    along with this program; if not, write to the Free Software
  17    Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA
  18*/
  19#include <linux/module.h>
  20#include <linux/moduleloader.h>
  21#include <linux/ftrace_event.h>
  22#include <linux/init.h>
  23#include <linux/kallsyms.h>
  24#include <linux/fs.h>
  25#include <linux/sysfs.h>
  26#include <linux/kernel.h>
  27#include <linux/slab.h>
  28#include <linux/vmalloc.h>
  29#include <linux/elf.h>
  30#include <linux/proc_fs.h>
  31#include <linux/seq_file.h>
  32#include <linux/syscalls.h>
  33#include <linux/fcntl.h>
  34#include <linux/rcupdate.h>
  35#include <linux/capability.h>
  36#include <linux/cpu.h>
  37#include <linux/moduleparam.h>
  38#include <linux/errno.h>
  39#include <linux/err.h>
  40#include <linux/vermagic.h>
  41#include <linux/notifier.h>
  42#include <linux/sched.h>
  43#include <linux/stop_machine.h>
  44#include <linux/device.h>
  45#include <linux/string.h>
  46#include <linux/mutex.h>
  47#include <linux/rculist.h>
  48#include <asm/uaccess.h>
  49#include <asm/cacheflush.h>
  50#include <asm/mmu_context.h>
  51#include <linux/license.h>
  52#include <asm/sections.h>
  53#include <linux/tracepoint.h>
  54#include <linux/ftrace.h>
  55#include <linux/async.h>
  56#include <linux/percpu.h>
  57#include <linux/kmemleak.h>
  58#include <linux/jump_label.h>
  59#include <linux/pfn.h>
  60
  61#define CREATE_TRACE_POINTS
  62#include <trace/events/module.h>
  63
  64#if 0
  65#define DEBUGP printk
  66#else
  67#define DEBUGP(fmt , a...)
  68#endif
  69
  70#ifndef ARCH_SHF_SMALL
  71#define ARCH_SHF_SMALL 0
  72#endif
  73
  74/*
  75 * Modules' sections will be aligned on page boundaries
  76 * to ensure complete separation of code and data, but
  77 * only when CONFIG_DEBUG_SET_MODULE_RONX=y
  78 */
  79#ifdef CONFIG_DEBUG_SET_MODULE_RONX
  80# define debug_align(X) ALIGN(X, PAGE_SIZE)
  81#else
  82# define debug_align(X) (X)
  83#endif
  84
  85/*
  86 * Given BASE and SIZE this macro calculates the number of pages the
  87 * memory regions occupies
  88 */
  89#define MOD_NUMBER_OF_PAGES(BASE, SIZE) (((SIZE) > 0) ?         \
  90                (PFN_DOWN((unsigned long)(BASE) + (SIZE) - 1) - \
  91                         PFN_DOWN((unsigned long)BASE) + 1)     \
  92                : (0UL))
  93
  94/* If this is set, the section belongs in the init part of the module */
  95#define INIT_OFFSET_MASK (1UL << (BITS_PER_LONG-1))
  96
  97/*
  98 * Mutex protects:
  99 * 1) List of modules (also safely readable with preempt_disable),
 100 * 2) module_use links,
 101 * 3) module_addr_min/module_addr_max.
 102 * (delete uses stop_machine/add uses RCU list operations). */
 103DEFINE_MUTEX(module_mutex);
 104EXPORT_SYMBOL_GPL(module_mutex);
 105static LIST_HEAD(modules);
 106#ifdef CONFIG_KGDB_KDB
 107struct list_head *kdb_modules = &modules; /* kdb needs the list of modules */
 108#endif /* CONFIG_KGDB_KDB */
 109
 110
 111/* Block module loading/unloading? */
 112int modules_disabled = 0;
 113
 114/* Waiting for a module to finish initializing? */
 115static DECLARE_WAIT_QUEUE_HEAD(module_wq);
 116
 117static BLOCKING_NOTIFIER_HEAD(module_notify_list);
 118
 119/* Bounds of module allocation, for speeding __module_address.
 120 * Protected by module_mutex. */
 121static unsigned long module_addr_min = -1UL, module_addr_max = 0;
 122
 123int register_module_notifier(struct notifier_block * nb)
 124{
 125        return blocking_notifier_chain_register(&module_notify_list, nb);
 126}
 127EXPORT_SYMBOL(register_module_notifier);
 128
 129int unregister_module_notifier(struct notifier_block * nb)
 130{
 131        return blocking_notifier_chain_unregister(&module_notify_list, nb);
 132}
 133EXPORT_SYMBOL(unregister_module_notifier);
 134
 135struct load_info {
 136        Elf_Ehdr *hdr;
 137        unsigned long len;
 138        Elf_Shdr *sechdrs;
 139        char *secstrings, *strtab;
 140        unsigned long *strmap;
 141        unsigned long symoffs, stroffs;
 142        struct _ddebug *debug;
 143        unsigned int num_debug;
 144        struct {
 145                unsigned int sym, str, mod, vers, info, pcpu;
 146        } index;
 147};
 148
 149/* We require a truly strong try_module_get(): 0 means failure due to
 150   ongoing or failed initialization etc. */
 151static inline int strong_try_module_get(struct module *mod)
 152{
 153        if (mod && mod->state == MODULE_STATE_COMING)
 154                return -EBUSY;
 155        if (try_module_get(mod))
 156                return 0;
 157        else
 158                return -ENOENT;
 159}
 160
 161static inline void add_taint_module(struct module *mod, unsigned flag)
 162{
 163        add_taint(flag);
 164        mod->taints |= (1U << flag);
 165}
 166
 167/*
 168 * A thread that wants to hold a reference to a module only while it
 169 * is running can call this to safely exit.  nfsd and lockd use this.
 170 */
 171void __module_put_and_exit(struct module *mod, long code)
 172{
 173        module_put(mod);
 174        do_exit(code);
 175}
 176EXPORT_SYMBOL(__module_put_and_exit);
 177
 178/* Find a module section: 0 means not found. */
 179static unsigned int find_sec(const struct load_info *info, const char *name)
 180{
 181        unsigned int i;
 182
 183        for (i = 1; i < info->hdr->e_shnum; i++) {
 184                Elf_Shdr *shdr = &info->sechdrs[i];
 185                /* Alloc bit cleared means "ignore it." */
 186                if ((shdr->sh_flags & SHF_ALLOC)
 187                    && strcmp(info->secstrings + shdr->sh_name, name) == 0)
 188                        return i;
 189        }
 190        return 0;
 191}
 192
 193/* Find a module section, or NULL. */
 194static void *section_addr(const struct load_info *info, const char *name)
 195{
 196        /* Section 0 has sh_addr 0. */
 197        return (void *)info->sechdrs[find_sec(info, name)].sh_addr;
 198}
 199
 200/* Find a module section, or NULL.  Fill in number of "objects" in section. */
 201static void *section_objs(const struct load_info *info,
 202                          const char *name,
 203                          size_t object_size,
 204                          unsigned int *num)
 205{
 206        unsigned int sec = find_sec(info, name);
 207
 208        /* Section 0 has sh_addr 0 and sh_size 0. */
 209        *num = info->sechdrs[sec].sh_size / object_size;
 210        return (void *)info->sechdrs[sec].sh_addr;
 211}
 212
 213/* Provided by the linker */
 214extern const struct kernel_symbol __start___ksymtab[];
 215extern const struct kernel_symbol __stop___ksymtab[];
 216extern const struct kernel_symbol __start___ksymtab_gpl[];
 217extern const struct kernel_symbol __stop___ksymtab_gpl[];
 218extern const struct kernel_symbol __start___ksymtab_gpl_future[];
 219extern const struct kernel_symbol __stop___ksymtab_gpl_future[];
 220extern const unsigned long __start___kcrctab[];
 221extern const unsigned long __start___kcrctab_gpl[];
 222extern const unsigned long __start___kcrctab_gpl_future[];
 223#ifdef CONFIG_UNUSED_SYMBOLS
 224extern const struct kernel_symbol __start___ksymtab_unused[];
 225extern const struct kernel_symbol __stop___ksymtab_unused[];
 226extern const struct kernel_symbol __start___ksymtab_unused_gpl[];
 227extern const struct kernel_symbol __stop___ksymtab_unused_gpl[];
 228extern const unsigned long __start___kcrctab_unused[];
 229extern const unsigned long __start___kcrctab_unused_gpl[];
 230#endif
 231
 232#ifndef CONFIG_MODVERSIONS
 233#define symversion(base, idx) NULL
 234#else
 235#define symversion(base, idx) ((base != NULL) ? ((base) + (idx)) : NULL)
 236#endif
 237
 238static bool each_symbol_in_section(const struct symsearch *arr,
 239                                   unsigned int arrsize,
 240                                   struct module *owner,
 241                                   bool (*fn)(const struct symsearch *syms,
 242                                              struct module *owner,
 243                                              unsigned int symnum, void *data),
 244                                   void *data)
 245{
 246        unsigned int i, j;
 247
 248        for (j = 0; j < arrsize; j++) {
 249                for (i = 0; i < arr[j].stop - arr[j].start; i++)
 250                        if (fn(&arr[j], owner, i, data))
 251                                return true;
 252        }
 253
 254        return false;
 255}
 256
 257/* Returns true as soon as fn returns true, otherwise false. */
 258bool each_symbol(bool (*fn)(const struct symsearch *arr, struct module *owner,
 259                            unsigned int symnum, void *data), void *data)
 260{
 261        struct module *mod;
 262        static const struct symsearch arr[] = {
 263                { __start___ksymtab, __stop___ksymtab, __start___kcrctab,
 264                  NOT_GPL_ONLY, false },
 265                { __start___ksymtab_gpl, __stop___ksymtab_gpl,
 266                  __start___kcrctab_gpl,
 267                  GPL_ONLY, false },
 268                { __start___ksymtab_gpl_future, __stop___ksymtab_gpl_future,
 269                  __start___kcrctab_gpl_future,
 270                  WILL_BE_GPL_ONLY, false },
 271#ifdef CONFIG_UNUSED_SYMBOLS
 272                { __start___ksymtab_unused, __stop___ksymtab_unused,
 273                  __start___kcrctab_unused,
 274                  NOT_GPL_ONLY, true },
 275                { __start___ksymtab_unused_gpl, __stop___ksymtab_unused_gpl,
 276                  __start___kcrctab_unused_gpl,
 277                  GPL_ONLY, true },
 278#endif
 279        };
 280
 281        if (each_symbol_in_section(arr, ARRAY_SIZE(arr), NULL, fn, data))
 282                return true;
 283
 284        list_for_each_entry_rcu(mod, &modules, list) {
 285                struct symsearch arr[] = {
 286                        { mod->syms, mod->syms + mod->num_syms, mod->crcs,
 287                          NOT_GPL_ONLY, false },
 288                        { mod->gpl_syms, mod->gpl_syms + mod->num_gpl_syms,
 289                          mod->gpl_crcs,
 290                          GPL_ONLY, false },
 291                        { mod->gpl_future_syms,
 292                          mod->gpl_future_syms + mod->num_gpl_future_syms,
 293                          mod->gpl_future_crcs,
 294                          WILL_BE_GPL_ONLY, false },
 295#ifdef CONFIG_UNUSED_SYMBOLS
 296                        { mod->unused_syms,
 297                          mod->unused_syms + mod->num_unused_syms,
 298                          mod->unused_crcs,
 299                          NOT_GPL_ONLY, true },
 300                        { mod->unused_gpl_syms,
 301                          mod->unused_gpl_syms + mod->num_unused_gpl_syms,
 302                          mod->unused_gpl_crcs,
 303                          GPL_ONLY, true },
 304#endif
 305                };
 306
 307                if (each_symbol_in_section(arr, ARRAY_SIZE(arr), mod, fn, data))
 308                        return true;
 309        }
 310        return false;
 311}
 312EXPORT_SYMBOL_GPL(each_symbol);
 313
 314struct find_symbol_arg {
 315        /* Input */
 316        const char *name;
 317        bool gplok;
 318        bool warn;
 319
 320        /* Output */
 321        struct module *owner;
 322        const unsigned long *crc;
 323        const struct kernel_symbol *sym;
 324};
 325
 326static bool find_symbol_in_section(const struct symsearch *syms,
 327                                   struct module *owner,
 328                                   unsigned int symnum, void *data)
 329{
 330        struct find_symbol_arg *fsa = data;
 331
 332        if (strcmp(syms->start[symnum].name, fsa->name) != 0)
 333                return false;
 334
 335        if (!fsa->gplok) {
 336                if (syms->licence == GPL_ONLY)
 337                        return false;
 338                if (syms->licence == WILL_BE_GPL_ONLY && fsa->warn) {
 339                        printk(KERN_WARNING "Symbol %s is being used "
 340                               "by a non-GPL module, which will not "
 341                               "be allowed in the future\n", fsa->name);
 342                        printk(KERN_WARNING "Please see the file "
 343                               "Documentation/feature-removal-schedule.txt "
 344                               "in the kernel source tree for more details.\n");
 345                }
 346        }
 347
 348#ifdef CONFIG_UNUSED_SYMBOLS
 349        if (syms->unused && fsa->warn) {
 350                printk(KERN_WARNING "Symbol %s is marked as UNUSED, "
 351                       "however this module is using it.\n", fsa->name);
 352                printk(KERN_WARNING
 353                       "This symbol will go away in the future.\n");
 354                printk(KERN_WARNING
 355                       "Please evalute if this is the right api to use and if "
 356                       "it really is, submit a report the linux kernel "
 357                       "mailinglist together with submitting your code for "
 358                       "inclusion.\n");
 359        }
 360#endif
 361
 362        fsa->owner = owner;
 363        fsa->crc = symversion(syms->crcs, symnum);
 364        fsa->sym = &syms->start[symnum];
 365        return true;
 366}
 367
 368/* Find a symbol and return it, along with, (optional) crc and
 369 * (optional) module which owns it.  Needs preempt disabled or module_mutex. */
 370const struct kernel_symbol *find_symbol(const char *name,
 371                                        struct module **owner,
 372                                        const unsigned long **crc,
 373                                        bool gplok,
 374                                        bool warn)
 375{
 376        struct find_symbol_arg fsa;
 377
 378        fsa.name = name;
 379        fsa.gplok = gplok;
 380        fsa.warn = warn;
 381
 382        if (each_symbol(find_symbol_in_section, &fsa)) {
 383                if (owner)
 384                        *owner = fsa.owner;
 385                if (crc)
 386                        *crc = fsa.crc;
 387                return fsa.sym;
 388        }
 389
 390        DEBUGP("Failed to find symbol %s\n", name);
 391        return NULL;
 392}
 393EXPORT_SYMBOL_GPL(find_symbol);
 394
 395/* Search for module by name: must hold module_mutex. */
 396struct module *find_module(const char *name)
 397{
 398        struct module *mod;
 399
 400        list_for_each_entry(mod, &modules, list) {
 401                if (strcmp(mod->name, name) == 0)
 402                        return mod;
 403        }
 404        return NULL;
 405}
 406EXPORT_SYMBOL_GPL(find_module);
 407
 408#ifdef CONFIG_SMP
 409
 410static inline void __percpu *mod_percpu(struct module *mod)
 411{
 412        return mod->percpu;
 413}
 414
 415static int percpu_modalloc(struct module *mod,
 416                           unsigned long size, unsigned long align)
 417{
 418        if (align > PAGE_SIZE) {
 419                printk(KERN_WARNING "%s: per-cpu alignment %li > %li\n",
 420                       mod->name, align, PAGE_SIZE);
 421                align = PAGE_SIZE;
 422        }
 423
 424        mod->percpu = __alloc_reserved_percpu(size, align);
 425        if (!mod->percpu) {
 426                printk(KERN_WARNING
 427                       "%s: Could not allocate %lu bytes percpu data\n",
 428                       mod->name, size);
 429                return -ENOMEM;
 430        }
 431        mod->percpu_size = size;
 432        return 0;
 433}
 434
 435static void percpu_modfree(struct module *mod)
 436{
 437        free_percpu(mod->percpu);
 438}
 439
 440static unsigned int find_pcpusec(struct load_info *info)
 441{
 442        return find_sec(info, ".data..percpu");
 443}
 444
 445static void percpu_modcopy(struct module *mod,
 446                           const void *from, unsigned long size)
 447{
 448        int cpu;
 449
 450        for_each_possible_cpu(cpu)
 451                memcpy(per_cpu_ptr(mod->percpu, cpu), from, size);
 452}
 453
 454/**
 455 * is_module_percpu_address - test whether address is from module static percpu
 456 * @addr: address to test
 457 *
 458 * Test whether @addr belongs to module static percpu area.
 459 *
 460 * RETURNS:
 461 * %true if @addr is from module static percpu area
 462 */
 463bool is_module_percpu_address(unsigned long addr)
 464{
 465        struct module *mod;
 466        unsigned int cpu;
 467
 468        preempt_disable();
 469
 470        list_for_each_entry_rcu(mod, &modules, list) {
 471                if (!mod->percpu_size)
 472                        continue;
 473                for_each_possible_cpu(cpu) {
 474                        void *start = per_cpu_ptr(mod->percpu, cpu);
 475
 476                        if ((void *)addr >= start &&
 477                            (void *)addr < start + mod->percpu_size) {
 478                                preempt_enable();
 479                                return true;
 480                        }
 481                }
 482        }
 483
 484        preempt_enable();
 485        return false;
 486}
 487
 488#else /* ... !CONFIG_SMP */
 489
 490static inline void __percpu *mod_percpu(struct module *mod)
 491{
 492        return NULL;
 493}
 494static inline int percpu_modalloc(struct module *mod,
 495                                  unsigned long size, unsigned long align)
 496{
 497        return -ENOMEM;
 498}
 499static inline void percpu_modfree(struct module *mod)
 500{
 501}
 502static unsigned int find_pcpusec(struct load_info *info)
 503{
 504        return 0;
 505}
 506static inline void percpu_modcopy(struct module *mod,
 507                                  const void *from, unsigned long size)
 508{
 509        /* pcpusec should be 0, and size of that section should be 0. */
 510        BUG_ON(size != 0);
 511}
 512bool is_module_percpu_address(unsigned long addr)
 513{
 514        return false;
 515}
 516
 517#endif /* CONFIG_SMP */
 518
 519#define MODINFO_ATTR(field)     \
 520static void setup_modinfo_##field(struct module *mod, const char *s)  \
 521{                                                                     \
 522        mod->field = kstrdup(s, GFP_KERNEL);                          \
 523}                                                                     \
 524static ssize_t show_modinfo_##field(struct module_attribute *mattr,   \
 525                        struct module *mod, char *buffer)             \
 526{                                                                     \
 527        return sprintf(buffer, "%s\n", mod->field);                   \
 528}                                                                     \
 529static int modinfo_##field##_exists(struct module *mod)               \
 530{                                                                     \
 531        return mod->field != NULL;                                    \
 532}                                                                     \
 533static void free_modinfo_##field(struct module *mod)                  \
 534{                                                                     \
 535        kfree(mod->field);                                            \
 536        mod->field = NULL;                                            \
 537}                                                                     \
 538static struct module_attribute modinfo_##field = {                    \
 539        .attr = { .name = __stringify(field), .mode = 0444 },         \
 540        .show = show_modinfo_##field,                                 \
 541        .setup = setup_modinfo_##field,                               \
 542        .test = modinfo_##field##_exists,                             \
 543        .free = free_modinfo_##field,                                 \
 544};
 545
 546MODINFO_ATTR(version);
 547MODINFO_ATTR(srcversion);
 548
 549static char last_unloaded_module[MODULE_NAME_LEN+1];
 550
 551#ifdef CONFIG_MODULE_UNLOAD
 552
 553EXPORT_TRACEPOINT_SYMBOL(module_get);
 554
 555/* Init the unload section of the module. */
 556static int module_unload_init(struct module *mod)
 557{
 558        mod->refptr = alloc_percpu(struct module_ref);
 559        if (!mod->refptr)
 560                return -ENOMEM;
 561
 562        INIT_LIST_HEAD(&mod->source_list);
 563        INIT_LIST_HEAD(&mod->target_list);
 564
 565        /* Hold reference count during initialization. */
 566        __this_cpu_write(mod->refptr->incs, 1);
 567        /* Backwards compatibility macros put refcount during init. */
 568        mod->waiter = current;
 569
 570        return 0;
 571}
 572
 573/* Does a already use b? */
 574static int already_uses(struct module *a, struct module *b)
 575{
 576        struct module_use *use;
 577
 578        list_for_each_entry(use, &b->source_list, source_list) {
 579                if (use->source == a) {
 580                        DEBUGP("%s uses %s!\n", a->name, b->name);
 581                        return 1;
 582                }
 583        }
 584        DEBUGP("%s does not use %s!\n", a->name, b->name);
 585        return 0;
 586}
 587
 588/*
 589 * Module a uses b
 590 *  - we add 'a' as a "source", 'b' as a "target" of module use
 591 *  - the module_use is added to the list of 'b' sources (so
 592 *    'b' can walk the list to see who sourced them), and of 'a'
 593 *    targets (so 'a' can see what modules it targets).
 594 */
 595static int add_module_usage(struct module *a, struct module *b)
 596{
 597        struct module_use *use;
 598
 599        DEBUGP("Allocating new usage for %s.\n", a->name);
 600        use = kmalloc(sizeof(*use), GFP_ATOMIC);
 601        if (!use) {
 602                printk(KERN_WARNING "%s: out of memory loading\n", a->name);
 603                return -ENOMEM;
 604        }
 605
 606        use->source = a;
 607        use->target = b;
 608        list_add(&use->source_list, &b->source_list);
 609        list_add(&use->target_list, &a->target_list);
 610        return 0;
 611}
 612
 613/* Module a uses b: caller needs module_mutex() */
 614int ref_module(struct module *a, struct module *b)
 615{
 616        int err;
 617
 618        if (b == NULL || already_uses(a, b))
 619                return 0;
 620
 621        /* If module isn't available, we fail. */
 622        err = strong_try_module_get(b);
 623        if (err)
 624                return err;
 625
 626        err = add_module_usage(a, b);
 627        if (err) {
 628                module_put(b);
 629                return err;
 630        }
 631        return 0;
 632}
 633EXPORT_SYMBOL_GPL(ref_module);
 634
 635/* Clear the unload stuff of the module. */
 636static void module_unload_free(struct module *mod)
 637{
 638        struct module_use *use, *tmp;
 639
 640        mutex_lock(&module_mutex);
 641        list_for_each_entry_safe(use, tmp, &mod->target_list, target_list) {
 642                struct module *i = use->target;
 643                DEBUGP("%s unusing %s\n", mod->name, i->name);
 644                module_put(i);
 645                list_del(&use->source_list);
 646                list_del(&use->target_list);
 647                kfree(use);
 648        }
 649        mutex_unlock(&module_mutex);
 650
 651        free_percpu(mod->refptr);
 652}
 653
 654#ifdef CONFIG_MODULE_FORCE_UNLOAD
 655static inline int try_force_unload(unsigned int flags)
 656{
 657        int ret = (flags & O_TRUNC);
 658        if (ret)
 659                add_taint(TAINT_FORCED_RMMOD);
 660        return ret;
 661}
 662#else
 663static inline int try_force_unload(unsigned int flags)
 664{
 665        return 0;
 666}
 667#endif /* CONFIG_MODULE_FORCE_UNLOAD */
 668
 669struct stopref
 670{
 671        struct module *mod;
 672        int flags;
 673        int *forced;
 674};
 675
 676/* Whole machine is stopped with interrupts off when this runs. */
 677static int __try_stop_module(void *_sref)
 678{
 679        struct stopref *sref = _sref;
 680
 681        /* If it's not unused, quit unless we're forcing. */
 682        if (module_refcount(sref->mod) != 0) {
 683                if (!(*sref->forced = try_force_unload(sref->flags)))
 684                        return -EWOULDBLOCK;
 685        }
 686
 687        /* Mark it as dying. */
 688        sref->mod->state = MODULE_STATE_GOING;
 689        return 0;
 690}
 691
 692static int try_stop_module(struct module *mod, int flags, int *forced)
 693{
 694        if (flags & O_NONBLOCK) {
 695                struct stopref sref = { mod, flags, forced };
 696
 697                return stop_machine(__try_stop_module, &sref, NULL);
 698        } else {
 699                /* We don't need to stop the machine for this. */
 700                mod->state = MODULE_STATE_GOING;
 701                synchronize_sched();
 702                return 0;
 703        }
 704}
 705
 706unsigned int module_refcount(struct module *mod)
 707{
 708        unsigned int incs = 0, decs = 0;
 709        int cpu;
 710
 711        for_each_possible_cpu(cpu)
 712                decs += per_cpu_ptr(mod->refptr, cpu)->decs;
 713        /*
 714         * ensure the incs are added up after the decs.
 715         * module_put ensures incs are visible before decs with smp_wmb.
 716         *
 717         * This 2-count scheme avoids the situation where the refcount
 718         * for CPU0 is read, then CPU0 increments the module refcount,
 719         * then CPU1 drops that refcount, then the refcount for CPU1 is
 720         * read. We would record a decrement but not its corresponding
 721         * increment so we would see a low count (disaster).
 722         *
 723         * Rare situation? But module_refcount can be preempted, and we
 724         * might be tallying up 4096+ CPUs. So it is not impossible.
 725         */
 726        smp_rmb();
 727        for_each_possible_cpu(cpu)
 728                incs += per_cpu_ptr(mod->refptr, cpu)->incs;
 729        return incs - decs;
 730}
 731EXPORT_SYMBOL(module_refcount);
 732
 733/* This exists whether we can unload or not */
 734static void free_module(struct module *mod);
 735
 736static void wait_for_zero_refcount(struct module *mod)
 737{
 738        /* Since we might sleep for some time, release the mutex first */
 739        mutex_unlock(&module_mutex);
 740        for (;;) {
 741                DEBUGP("Looking at refcount...\n");
 742                set_current_state(TASK_UNINTERRUPTIBLE);
 743                if (module_refcount(mod) == 0)
 744                        break;
 745                schedule();
 746        }
 747        current->state = TASK_RUNNING;
 748        mutex_lock(&module_mutex);
 749}
 750
 751SYSCALL_DEFINE2(delete_module, const char __user *, name_user,
 752                unsigned int, flags)
 753{
 754        struct module *mod;
 755        char name[MODULE_NAME_LEN];
 756        int ret, forced = 0;
 757
 758        if (!capable(CAP_SYS_MODULE) || modules_disabled)
 759                return -EPERM;
 760
 761        if (strncpy_from_user(name, name_user, MODULE_NAME_LEN-1) < 0)
 762                return -EFAULT;
 763        name[MODULE_NAME_LEN-1] = '\0';
 764
 765        if (mutex_lock_interruptible(&module_mutex) != 0)
 766                return -EINTR;
 767
 768        mod = find_module(name);
 769        if (!mod) {
 770                ret = -ENOENT;
 771                goto out;
 772        }
 773
 774        if (!list_empty(&mod->source_list)) {
 775                /* Other modules depend on us: get rid of them first. */
 776                ret = -EWOULDBLOCK;
 777                goto out;
 778        }
 779
 780        /* Doing init or already dying? */
 781        if (mod->state != MODULE_STATE_LIVE) {
 782                /* FIXME: if (force), slam module count and wake up
 783                   waiter --RR */
 784                DEBUGP("%s already dying\n", mod->name);
 785                ret = -EBUSY;
 786                goto out;
 787        }
 788
 789        /* If it has an init func, it must have an exit func to unload */
 790        if (mod->init && !mod->exit) {
 791                forced = try_force_unload(flags);
 792                if (!forced) {
 793                        /* This module can't be removed */
 794                        ret = -EBUSY;
 795                        goto out;
 796                }
 797        }
 798
 799        /* Set this up before setting mod->state */
 800        mod->waiter = current;
 801
 802        /* Stop the machine so refcounts can't move and disable module. */
 803        ret = try_stop_module(mod, flags, &forced);
 804        if (ret != 0)
 805                goto out;
 806
 807        /* Never wait if forced. */
 808        if (!forced && module_refcount(mod) != 0)
 809                wait_for_zero_refcount(mod);
 810
 811        mutex_unlock(&module_mutex);
 812        /* Final destruction now noone is using it. */
 813        if (mod->exit != NULL)
 814                mod->exit();
 815        blocking_notifier_call_chain(&module_notify_list,
 816                                     MODULE_STATE_GOING, mod);
 817        async_synchronize_full();
 818
 819        /* Store the name of the last unloaded module for diagnostic purposes */
 820        strlcpy(last_unloaded_module, mod->name, sizeof(last_unloaded_module));
 821
 822        free_module(mod);
 823        return 0;
 824out:
 825        mutex_unlock(&module_mutex);
 826        return ret;
 827}
 828
 829static inline void print_unload_info(struct seq_file *m, struct module *mod)
 830{
 831        struct module_use *use;
 832        int printed_something = 0;
 833
 834        seq_printf(m, " %u ", module_refcount(mod));
 835
 836        /* Always include a trailing , so userspace can differentiate
 837           between this and the old multi-field proc format. */
 838        list_for_each_entry(use, &mod->source_list, source_list) {
 839                printed_something = 1;
 840                seq_printf(m, "%s,", use->source->name);
 841        }
 842
 843        if (mod->init != NULL && mod->exit == NULL) {
 844                printed_something = 1;
 845                seq_printf(m, "[permanent],");
 846        }
 847
 848        if (!printed_something)
 849                seq_printf(m, "-");
 850}
 851
 852void __symbol_put(const char *symbol)
 853{
 854        struct module *owner;
 855
 856        preempt_disable();
 857        if (!find_symbol(symbol, &owner, NULL, true, false))
 858                BUG();
 859        module_put(owner);
 860        preempt_enable();
 861}
 862EXPORT_SYMBOL(__symbol_put);
 863
 864/* Note this assumes addr is a function, which it currently always is. */
 865void symbol_put_addr(void *addr)
 866{
 867        struct module *modaddr;
 868        unsigned long a = (unsigned long)dereference_function_descriptor(addr);
 869
 870        if (core_kernel_text(a))
 871                return;
 872
 873        /* module_text_address is safe here: we're supposed to have reference
 874         * to module from symbol_get, so it can't go away. */
 875        modaddr = __module_text_address(a);
 876        BUG_ON(!modaddr);
 877        module_put(modaddr);
 878}
 879EXPORT_SYMBOL_GPL(symbol_put_addr);
 880
 881static ssize_t show_refcnt(struct module_attribute *mattr,
 882                           struct module *mod, char *buffer)
 883{
 884        return sprintf(buffer, "%u\n", module_refcount(mod));
 885}
 886
 887static struct module_attribute refcnt = {
 888        .attr = { .name = "refcnt", .mode = 0444 },
 889        .show = show_refcnt,
 890};
 891
 892void module_put(struct module *module)
 893{
 894        if (module) {
 895                preempt_disable();
 896                smp_wmb(); /* see comment in module_refcount */
 897                __this_cpu_inc(module->refptr->decs);
 898
 899                trace_module_put(module, _RET_IP_);
 900                /* Maybe they're waiting for us to drop reference? */
 901                if (unlikely(!module_is_live(module)))
 902                        wake_up_process(module->waiter);
 903                preempt_enable();
 904        }
 905}
 906EXPORT_SYMBOL(module_put);
 907
 908#else /* !CONFIG_MODULE_UNLOAD */
 909static inline void print_unload_info(struct seq_file *m, struct module *mod)
 910{
 911        /* We don't know the usage count, or what modules are using. */
 912        seq_printf(m, " - -");
 913}
 914
 915static inline void module_unload_free(struct module *mod)
 916{
 917}
 918
 919int ref_module(struct module *a, struct module *b)
 920{
 921        return strong_try_module_get(b);
 922}
 923EXPORT_SYMBOL_GPL(ref_module);
 924
 925static inline int module_unload_init(struct module *mod)
 926{
 927        return 0;
 928}
 929#endif /* CONFIG_MODULE_UNLOAD */
 930
 931static ssize_t show_initstate(struct module_attribute *mattr,
 932                           struct module *mod, char *buffer)
 933{
 934        const char *state = "unknown";
 935
 936        switch (mod->state) {
 937        case MODULE_STATE_LIVE:
 938                state = "live";
 939                break;
 940        case MODULE_STATE_COMING:
 941                state = "coming";
 942                break;
 943        case MODULE_STATE_GOING:
 944                state = "going";
 945                break;
 946        }
 947        return sprintf(buffer, "%s\n", state);
 948}
 949
 950static struct module_attribute initstate = {
 951        .attr = { .name = "initstate", .mode = 0444 },
 952        .show = show_initstate,
 953};
 954
 955static struct module_attribute *modinfo_attrs[] = {
 956        &modinfo_version,
 957        &modinfo_srcversion,
 958        &initstate,
 959#ifdef CONFIG_MODULE_UNLOAD
 960        &refcnt,
 961#endif
 962        NULL,
 963};
 964
 965static const char vermagic[] = VERMAGIC_STRING;
 966
 967static int try_to_force_load(struct module *mod, const char *reason)
 968{
 969#ifdef CONFIG_MODULE_FORCE_LOAD
 970        if (!test_taint(TAINT_FORCED_MODULE))
 971                printk(KERN_WARNING "%s: %s: kernel tainted.\n",
 972                       mod->name, reason);
 973        add_taint_module(mod, TAINT_FORCED_MODULE);
 974        return 0;
 975#else
 976        return -ENOEXEC;
 977#endif
 978}
 979
 980#ifdef CONFIG_MODVERSIONS
 981/* If the arch applies (non-zero) relocations to kernel kcrctab, unapply it. */
 982static unsigned long maybe_relocated(unsigned long crc,
 983                                     const struct module *crc_owner)
 984{
 985#ifdef ARCH_RELOCATES_KCRCTAB
 986        if (crc_owner == NULL)
 987                return crc - (unsigned long)reloc_start;
 988#endif
 989        return crc;
 990}
 991
 992static int check_version(Elf_Shdr *sechdrs,
 993                         unsigned int versindex,
 994                         const char *symname,
 995                         struct module *mod, 
 996                         const unsigned long *crc,
 997                         const struct module *crc_owner)
 998{
 999        unsigned int i, num_versions;
1000        struct modversion_info *versions;
1001
1002        /* Exporting module didn't supply crcs?  OK, we're already tainted. */
1003        if (!crc)
1004                return 1;
1005
1006        /* No versions at all?  modprobe --force does this. */
1007        if (versindex == 0)
1008                return try_to_force_load(mod, symname) == 0;
1009
1010        versions = (void *) sechdrs[versindex].sh_addr;
1011        num_versions = sechdrs[versindex].sh_size
1012                / sizeof(struct modversion_info);
1013
1014        for (i = 0; i < num_versions; i++) {
1015                if (strcmp(versions[i].name, symname) != 0)
1016                        continue;
1017
1018                if (versions[i].crc == maybe_relocated(*crc, crc_owner))
1019                        return 1;
1020                DEBUGP("Found checksum %lX vs module %lX\n",
1021                       maybe_relocated(*crc, crc_owner), versions[i].crc);
1022                goto bad_version;
1023        }
1024
1025        printk(KERN_WARNING "%s: no symbol version for %s\n",
1026               mod->name, symname);
1027        return 0;
1028
1029bad_version:
1030        printk("%s: disagrees about version of symbol %s\n",
1031               mod->name, symname);
1032        return 0;
1033}
1034
1035static inline int check_modstruct_version(Elf_Shdr *sechdrs,
1036                                          unsigned int versindex,
1037                                          struct module *mod)
1038{
1039        const unsigned long *crc;
1040
1041        /* Since this should be found in kernel (which can't be removed),
1042         * no locking is necessary. */
1043        if (!find_symbol(MODULE_SYMBOL_PREFIX "module_layout", NULL,
1044                         &crc, true, false))
1045                BUG();
1046        return check_version(sechdrs, versindex, "module_layout", mod, crc,
1047                             NULL);
1048}
1049
1050/* First part is kernel version, which we ignore if module has crcs. */
1051static inline int same_magic(const char *amagic, const char *bmagic,
1052                             bool has_crcs)
1053{
1054        if (has_crcs) {
1055                amagic += strcspn(amagic, " ");
1056                bmagic += strcspn(bmagic, " ");
1057        }
1058        return strcmp(amagic, bmagic) == 0;
1059}
1060#else
1061static inline int check_version(Elf_Shdr *sechdrs,
1062                                unsigned int versindex,
1063                                const char *symname,
1064                                struct module *mod, 
1065                                const unsigned long *crc,
1066                                const struct module *crc_owner)
1067{
1068        return 1;
1069}
1070
1071static inline int check_modstruct_version(Elf_Shdr *sechdrs,
1072                                          unsigned int versindex,
1073                                          struct module *mod)
1074{
1075        return 1;
1076}
1077
1078static inline int same_magic(const char *amagic, const char *bmagic,
1079                             bool has_crcs)
1080{
1081        return strcmp(amagic, bmagic) == 0;
1082}
1083#endif /* CONFIG_MODVERSIONS */
1084
1085/* Resolve a symbol for this module.  I.e. if we find one, record usage. */
1086static const struct kernel_symbol *resolve_symbol(struct module *mod,
1087                                                  const struct load_info *info,
1088                                                  const char *name,
1089                                                  char ownername[])
1090{
1091        struct module *owner;
1092        const struct kernel_symbol *sym;
1093        const unsigned long *crc;
1094        int err;
1095
1096        mutex_lock(&module_mutex);
1097        sym = find_symbol(name, &owner, &crc,
1098                          !(mod->taints & (1 << TAINT_PROPRIETARY_MODULE)), true);
1099        if (!sym)
1100                goto unlock;
1101
1102        if (!check_version(info->sechdrs, info->index.vers, name, mod, crc,
1103                           owner)) {
1104                sym = ERR_PTR(-EINVAL);
1105                goto getname;
1106        }
1107
1108        err = ref_module(mod, owner);
1109        if (err) {
1110                sym = ERR_PTR(err);
1111                goto getname;
1112        }
1113
1114getname:
1115        /* We must make copy under the lock if we failed to get ref. */
1116        strncpy(ownername, module_name(owner), MODULE_NAME_LEN);
1117unlock:
1118        mutex_unlock(&module_mutex);
1119        return sym;
1120}
1121
1122static const struct kernel_symbol *
1123resolve_symbol_wait(struct module *mod,
1124                    const struct load_info *info,
1125                    const char *name)
1126{
1127        const struct kernel_symbol *ksym;
1128        char owner[MODULE_NAME_LEN];
1129
1130        if (wait_event_interruptible_timeout(module_wq,
1131                        !IS_ERR(ksym = resolve_symbol(mod, info, name, owner))
1132                        || PTR_ERR(ksym) != -EBUSY,
1133                                             30 * HZ) <= 0) {
1134                printk(KERN_WARNING "%s: gave up waiting for init of module %s.\n",
1135                       mod->name, owner);
1136        }
1137        return ksym;
1138}
1139
1140/*
1141 * /sys/module/foo/sections stuff
1142 * J. Corbet <corbet@lwn.net>
1143 */
1144#ifdef CONFIG_SYSFS
1145
1146#ifdef CONFIG_KALLSYMS
1147static inline bool sect_empty(const Elf_Shdr *sect)
1148{
1149        return !(sect->sh_flags & SHF_ALLOC) || sect->sh_size == 0;
1150}
1151
1152struct module_sect_attr
1153{
1154        struct module_attribute mattr;
1155        char *name;
1156        unsigned long address;
1157};
1158
1159struct module_sect_attrs
1160{
1161        struct attribute_group grp;
1162        unsigned int nsections;
1163        struct module_sect_attr attrs[0];
1164};
1165
1166static ssize_t module_sect_show(struct module_attribute *mattr,
1167                                struct module *mod, char *buf)
1168{
1169        struct module_sect_attr *sattr =
1170                container_of(mattr, struct module_sect_attr, mattr);
1171        return sprintf(buf, "0x%lx\n", sattr->address);
1172}
1173
1174static void free_sect_attrs(struct module_sect_attrs *sect_attrs)
1175{
1176        unsigned int section;
1177
1178        for (section = 0; section < sect_attrs->nsections; section++)
1179                kfree(sect_attrs->attrs[section].name);
1180        kfree(sect_attrs);
1181}
1182
1183static void add_sect_attrs(struct module *mod, const struct load_info *info)
1184{
1185        unsigned int nloaded = 0, i, size[2];
1186        struct module_sect_attrs *sect_attrs;
1187        struct module_sect_attr *sattr;
1188        struct attribute **gattr;
1189
1190        /* Count loaded sections and allocate structures */
1191        for (i = 0; i < info->hdr->e_shnum; i++)
1192                if (!sect_empty(&info->sechdrs[i]))
1193                        nloaded++;
1194        size[0] = ALIGN(sizeof(*sect_attrs)
1195                        + nloaded * sizeof(sect_attrs->attrs[0]),
1196                        sizeof(sect_attrs->grp.attrs[0]));
1197        size[1] = (nloaded + 1) * sizeof(sect_attrs->grp.attrs[0]);
1198        sect_attrs = kzalloc(size[0] + size[1], GFP_KERNEL);
1199        if (sect_attrs == NULL)
1200                return;
1201
1202        /* Setup section attributes. */
1203        sect_attrs->grp.name = "sections";
1204        sect_attrs->grp.attrs = (void *)sect_attrs + size[0];
1205
1206        sect_attrs->nsections = 0;
1207        sattr = &sect_attrs->attrs[0];
1208        gattr = &sect_attrs->grp.attrs[0];
1209        for (i = 0; i < info->hdr->e_shnum; i++) {
1210                Elf_Shdr *sec = &info->sechdrs[i];
1211                if (sect_empty(sec))
1212                        continue;
1213                sattr->address = sec->sh_addr;
1214                sattr->name = kstrdup(info->secstrings + sec->sh_name,
1215                                        GFP_KERNEL);
1216                if (sattr->name == NULL)
1217                        goto out;
1218                sect_attrs->nsections++;
1219                sysfs_attr_init(&sattr->mattr.attr);
1220                sattr->mattr.show = module_sect_show;
1221                sattr->mattr.store = NULL;
1222                sattr->mattr.attr.name = sattr->name;
1223                sattr->mattr.attr.mode = S_IRUGO;
1224                *(gattr++) = &(sattr++)->mattr.attr;
1225        }
1226        *gattr = NULL;
1227
1228        if (sysfs_create_group(&mod->mkobj.kobj, &sect_attrs->grp))
1229                goto out;
1230
1231        mod->sect_attrs = sect_attrs;
1232        return;
1233  out:
1234        free_sect_attrs(sect_attrs);
1235}
1236
1237static void remove_sect_attrs(struct module *mod)
1238{
1239        if (mod->sect_attrs) {
1240                sysfs_remove_group(&mod->mkobj.kobj,
1241                                   &mod->sect_attrs->grp);
1242                /* We are positive that no one is using any sect attrs
1243                 * at this point.  Deallocate immediately. */
1244                free_sect_attrs(mod->sect_attrs);
1245                mod->sect_attrs = NULL;
1246        }
1247}
1248
1249/*
1250 * /sys/module/foo/notes/.section.name gives contents of SHT_NOTE sections.
1251 */
1252
1253struct module_notes_attrs {
1254        struct kobject *dir;
1255        unsigned int notes;
1256        struct bin_attribute attrs[0];
1257};
1258
1259static ssize_t module_notes_read(struct file *filp, struct kobject *kobj,
1260                                 struct bin_attribute *bin_attr,
1261                                 char *buf, loff_t pos, size_t count)
1262{
1263        /*
1264         * The caller checked the pos and count against our size.
1265         */
1266        memcpy(buf, bin_attr->private + pos, count);
1267        return count;
1268}
1269
1270static void free_notes_attrs(struct module_notes_attrs *notes_attrs,
1271                             unsigned int i)
1272{
1273        if (notes_attrs->dir) {
1274                while (i-- > 0)
1275                        sysfs_remove_bin_file(notes_attrs->dir,
1276                                              &notes_attrs->attrs[i]);
1277                kobject_put(notes_attrs->dir);
1278        }
1279        kfree(notes_attrs);
1280}
1281
1282static void add_notes_attrs(struct module *mod, const struct load_info *info)
1283{
1284        unsigned int notes, loaded, i;
1285        struct module_notes_attrs *notes_attrs;
1286        struct bin_attribute *nattr;
1287
1288        /* failed to create section attributes, so can't create notes */
1289        if (!mod->sect_attrs)
1290                return;
1291
1292        /* Count notes sections and allocate structures.  */
1293        notes = 0;
1294        for (i = 0; i < info->hdr->e_shnum; i++)
1295                if (!sect_empty(&info->sechdrs[i]) &&
1296                    (info->sechdrs[i].sh_type == SHT_NOTE))
1297                        ++notes;
1298
1299        if (notes == 0)
1300                return;
1301
1302        notes_attrs = kzalloc(sizeof(*notes_attrs)
1303                              + notes * sizeof(notes_attrs->attrs[0]),
1304                              GFP_KERNEL);
1305        if (notes_attrs == NULL)
1306                return;
1307
1308        notes_attrs->notes = notes;
1309        nattr = &notes_attrs->attrs[0];
1310        for (loaded = i = 0; i < info->hdr->e_shnum; ++i) {
1311                if (sect_empty(&info->sechdrs[i]))
1312                        continue;
1313                if (info->sechdrs[i].sh_type == SHT_NOTE) {
1314                        sysfs_bin_attr_init(nattr);
1315                        nattr->attr.name = mod->sect_attrs->attrs[loaded].name;
1316                        nattr->attr.mode = S_IRUGO;
1317                        nattr->size = info->sechdrs[i].sh_size;
1318                        nattr->private = (void *) info->sechdrs[i].sh_addr;
1319                        nattr->read = module_notes_read;
1320                        ++nattr;
1321                }
1322                ++loaded;
1323        }
1324
1325        notes_attrs->dir = kobject_create_and_add("notes", &mod->mkobj.kobj);
1326        if (!notes_attrs->dir)
1327                goto out;
1328
1329        for (i = 0; i < notes; ++i)
1330                if (sysfs_create_bin_file(notes_attrs->dir,
1331                                          &notes_attrs->attrs[i]))
1332                        goto out;
1333
1334        mod->notes_attrs = notes_attrs;
1335        return;
1336
1337  out:
1338        free_notes_attrs(notes_attrs, i);
1339}
1340
1341static void remove_notes_attrs(struct module *mod)
1342{
1343        if (mod->notes_attrs)
1344                free_notes_attrs(mod->notes_attrs, mod->notes_attrs->notes);
1345}
1346
1347#else
1348
1349static inline void add_sect_attrs(struct module *mod,
1350                                  const struct load_info *info)
1351{
1352}
1353
1354static inline void remove_sect_attrs(struct module *mod)
1355{
1356}
1357
1358static inline void add_notes_attrs(struct module *mod,
1359                                   const struct load_info *info)
1360{
1361}
1362
1363static inline void remove_notes_attrs(struct module *mod)
1364{
1365}
1366#endif /* CONFIG_KALLSYMS */
1367
1368static void add_usage_links(struct module *mod)
1369{
1370#ifdef CONFIG_MODULE_UNLOAD
1371        struct module_use *use;
1372        int nowarn;
1373
1374        mutex_lock(&module_mutex);
1375        list_for_each_entry(use, &mod->target_list, target_list) {
1376                nowarn = sysfs_create_link(use->target->holders_dir,
1377                                           &mod->mkobj.kobj, mod->name);
1378        }
1379        mutex_unlock(&module_mutex);
1380#endif
1381}
1382
1383static void del_usage_links(struct module *mod)
1384{
1385#ifdef CONFIG_MODULE_UNLOAD
1386        struct module_use *use;
1387
1388        mutex_lock(&module_mutex);
1389        list_for_each_entry(use, &mod->target_list, target_list)
1390                sysfs_remove_link(use->target->holders_dir, mod->name);
1391        mutex_unlock(&module_mutex);
1392#endif
1393}
1394
1395static int module_add_modinfo_attrs(struct module *mod)
1396{
1397        struct module_attribute *attr;
1398        struct module_attribute *temp_attr;
1399        int error = 0;
1400        int i;
1401
1402        mod->modinfo_attrs = kzalloc((sizeof(struct module_attribute) *
1403                                        (ARRAY_SIZE(modinfo_attrs) + 1)),
1404                                        GFP_KERNEL);
1405        if (!mod->modinfo_attrs)
1406                return -ENOMEM;
1407
1408        temp_attr = mod->modinfo_attrs;
1409        for (i = 0; (attr = modinfo_attrs[i]) && !error; i++) {
1410                if (!attr->test ||
1411                    (attr->test && attr->test(mod))) {
1412                        memcpy(temp_attr, attr, sizeof(*temp_attr));
1413                        sysfs_attr_init(&temp_attr->attr);
1414                        error = sysfs_create_file(&mod->mkobj.kobj,&temp_attr->attr);
1415                        ++temp_attr;
1416                }
1417        }
1418        return error;
1419}
1420
1421static void module_remove_modinfo_attrs(struct module *mod)
1422{
1423        struct module_attribute *attr;
1424        int i;
1425
1426        for (i = 0; (attr = &mod->modinfo_attrs[i]); i++) {
1427                /* pick a field to test for end of list */
1428                if (!attr->attr.name)
1429                        break;
1430                sysfs_remove_file(&mod->mkobj.kobj,&attr->attr);
1431                if (attr->free)
1432                        attr->free(mod);
1433        }
1434        kfree(mod->modinfo_attrs);
1435}
1436
1437static int mod_sysfs_init(struct module *mod)
1438{
1439        int err;
1440        struct kobject *kobj;
1441
1442        if (!module_sysfs_initialized) {
1443                printk(KERN_ERR "%s: module sysfs not initialized\n",
1444                       mod->name);
1445                err = -EINVAL;
1446                goto out;
1447        }
1448
1449        kobj = kset_find_obj(module_kset, mod->name);
1450        if (kobj) {
1451                printk(KERN_ERR "%s: module is already loaded\n", mod->name);
1452                kobject_put(kobj);
1453                err = -EINVAL;
1454                goto out;
1455        }
1456
1457        mod->mkobj.mod = mod;
1458
1459        memset(&mod->mkobj.kobj, 0, sizeof(mod->mkobj.kobj));
1460        mod->mkobj.kobj.kset = module_kset;
1461        err = kobject_init_and_add(&mod->mkobj.kobj, &module_ktype, NULL,
1462                                   "%s", mod->name);
1463        if (err)
1464                kobject_put(&mod->mkobj.kobj);
1465
1466        /* delay uevent until full sysfs population */
1467out:
1468        return err;
1469}
1470
1471static int mod_sysfs_setup(struct module *mod,
1472                           const struct load_info *info,
1473                           struct kernel_param *kparam,
1474                           unsigned int num_params)
1475{
1476        int err;
1477
1478        err = mod_sysfs_init(mod);
1479        if (err)
1480                goto out;
1481
1482        mod->holders_dir = kobject_create_and_add("holders", &mod->mkobj.kobj);
1483        if (!mod->holders_dir) {
1484                err = -ENOMEM;
1485                goto out_unreg;
1486        }
1487
1488        err = module_param_sysfs_setup(mod, kparam, num_params);
1489        if (err)
1490                goto out_unreg_holders;
1491
1492        err = module_add_modinfo_attrs(mod);
1493        if (err)
1494                goto out_unreg_param;
1495
1496        add_usage_links(mod);
1497        add_sect_attrs(mod, info);
1498        add_notes_attrs(mod, info);
1499
1500        kobject_uevent(&mod->mkobj.kobj, KOBJ_ADD);
1501        return 0;
1502
1503out_unreg_param:
1504        module_param_sysfs_remove(mod);
1505out_unreg_holders:
1506        kobject_put(mod->holders_dir);
1507out_unreg:
1508        kobject_put(&mod->mkobj.kobj);
1509out:
1510        return err;
1511}
1512
1513static void mod_sysfs_fini(struct module *mod)
1514{
1515        remove_notes_attrs(mod);
1516        remove_sect_attrs(mod);
1517        kobject_put(&mod->mkobj.kobj);
1518}
1519
1520#else /* !CONFIG_SYSFS */
1521
1522static int mod_sysfs_setup(struct module *mod,
1523                           const struct load_info *info,
1524                           struct kernel_param *kparam,
1525                           unsigned int num_params)
1526{
1527        return 0;
1528}
1529
1530static void mod_sysfs_fini(struct module *mod)
1531{
1532}
1533
1534static void module_remove_modinfo_attrs(struct module *mod)
1535{
1536}
1537
1538static void del_usage_links(struct module *mod)
1539{
1540}
1541
1542#endif /* CONFIG_SYSFS */
1543
1544static void mod_sysfs_teardown(struct module *mod)
1545{
1546        del_usage_links(mod);
1547        module_remove_modinfo_attrs(mod);
1548        module_param_sysfs_remove(mod);
1549        kobject_put(mod->mkobj.drivers_dir);
1550        kobject_put(mod->holders_dir);
1551        mod_sysfs_fini(mod);
1552}
1553
1554/*
1555 * unlink the module with the whole machine is stopped with interrupts off
1556 * - this defends against kallsyms not taking locks
1557 */
1558static int __unlink_module(void *_mod)
1559{
1560        struct module *mod = _mod;
1561        list_del(&mod->list);
1562        module_bug_cleanup(mod);
1563        return 0;
1564}
1565
1566#ifdef CONFIG_DEBUG_SET_MODULE_RONX
1567/*
1568 * LKM RO/NX protection: protect module's text/ro-data
1569 * from modification and any data from execution.
1570 */
1571void set_page_attributes(void *start, void *end, int (*set)(unsigned long start, int num_pages))
1572{
1573        unsigned long begin_pfn = PFN_DOWN((unsigned long)start);
1574        unsigned long end_pfn = PFN_DOWN((unsigned long)end);
1575
1576        if (end_pfn > begin_pfn)
1577                set(begin_pfn << PAGE_SHIFT, end_pfn - begin_pfn);
1578}
1579
1580static void set_section_ro_nx(void *base,
1581                        unsigned long text_size,
1582                        unsigned long ro_size,
1583                        unsigned long total_size)
1584{
1585        /* begin and end PFNs of the current subsection */
1586        unsigned long begin_pfn;
1587        unsigned long end_pfn;
1588
1589        /*
1590         * Set RO for module text and RO-data:
1591         * - Always protect first page.
1592         * - Do not protect last partial page.
1593         */
1594        if (ro_size > 0)
1595                set_page_attributes(base, base + ro_size, set_memory_ro);
1596
1597        /*
1598         * Set NX permissions for module data:
1599         * - Do not protect first partial page.
1600         * - Always protect last page.
1601         */
1602        if (total_size > text_size) {
1603                begin_pfn = PFN_UP((unsigned long)base + text_size);
1604                end_pfn = PFN_UP((unsigned long)base + total_size);
1605                if (end_pfn > begin_pfn)
1606                        set_memory_nx(begin_pfn << PAGE_SHIFT, end_pfn - begin_pfn);
1607        }
1608}
1609
1610/* Setting memory back to RW+NX before releasing it */
1611void unset_section_ro_nx(struct module *mod, void *module_region)
1612{
1613        unsigned long total_pages;
1614
1615        if (mod->module_core == module_region) {
1616                /* Set core as NX+RW */
1617                total_pages = MOD_NUMBER_OF_PAGES(mod->module_core, mod->core_size);
1618                set_memory_nx((unsigned long)mod->module_core, total_pages);
1619                set_memory_rw((unsigned long)mod->module_core, total_pages);
1620
1621        } else if (mod->module_init == module_region) {
1622                /* Set init as NX+RW */
1623                total_pages = MOD_NUMBER_OF_PAGES(mod->module_init, mod->init_size);
1624                set_memory_nx((unsigned long)mod->module_init, total_pages);
1625                set_memory_rw((unsigned long)mod->module_init, total_pages);
1626        }
1627}
1628
1629/* Iterate through all modules and set each module's text as RW */
1630void set_all_modules_text_rw()
1631{
1632        struct module *mod;
1633
1634        mutex_lock(&module_mutex);
1635        list_for_each_entry_rcu(mod, &modules, list) {
1636                if ((mod->module_core) && (mod->core_text_size)) {
1637                        set_page_attributes(mod->module_core,
1638                                                mod->module_core + mod->core_text_size,
1639                                                set_memory_rw);
1640                }
1641                if ((mod->module_init) && (mod->init_text_size)) {
1642                        set_page_attributes(mod->module_init,
1643                                                mod->module_init + mod->init_text_size,
1644                                                set_memory_rw);
1645                }
1646        }
1647        mutex_unlock(&module_mutex);
1648}
1649
1650/* Iterate through all modules and set each module's text as RO */
1651void set_all_modules_text_ro()
1652{
1653        struct module *mod;
1654
1655        mutex_lock(&module_mutex);
1656        list_for_each_entry_rcu(mod, &modules, list) {
1657                if ((mod->module_core) && (mod->core_text_size)) {
1658                        set_page_attributes(mod->module_core,
1659                                                mod->module_core + mod->core_text_size,
1660                                                set_memory_ro);
1661                }
1662                if ((mod->module_init) && (mod->init_text_size)) {
1663                        set_page_attributes(mod->module_init,
1664                                                mod->module_init + mod->init_text_size,
1665                                                set_memory_ro);
1666                }
1667        }
1668        mutex_unlock(&module_mutex);
1669}
1670#else
1671static inline void set_section_ro_nx(void *base, unsigned long text_size, unsigned long ro_size, unsigned long total_size) { }
1672static inline void unset_section_ro_nx(struct module *mod, void *module_region) { }
1673#endif
1674
1675/* Free a module, remove from lists, etc. */
1676static void free_module(struct module *mod)
1677{
1678        trace_module_free(mod);
1679
1680        /* Delete from various lists */
1681        mutex_lock(&module_mutex);
1682        stop_machine(__unlink_module, mod, NULL);
1683        mutex_unlock(&module_mutex);
1684        mod_sysfs_teardown(mod);
1685
1686        /* Remove dynamic debug info */
1687        ddebug_remove_module(mod->name);
1688
1689        /* Arch-specific cleanup. */
1690        module_arch_cleanup(mod);
1691
1692        /* Module unload stuff */
1693        module_unload_free(mod);
1694
1695        /* Free any allocated parameters. */
1696        destroy_params(mod->kp, mod->num_kp);
1697
1698        /* This may be NULL, but that's OK */
1699        unset_section_ro_nx(mod, mod->module_init);
1700        module_free(mod, mod->module_init);
1701        kfree(mod->args);
1702        percpu_modfree(mod);
1703
1704        /* Free lock-classes: */
1705        lockdep_free_key_range(mod->module_core, mod->core_size);
1706
1707        /* Finally, free the core (containing the module structure) */
1708        unset_section_ro_nx(mod, mod->module_core);
1709        module_free(mod, mod->module_core);
1710
1711#ifdef CONFIG_MPU
1712        update_protections(current->mm);
1713#endif
1714}
1715
1716void *__symbol_get(const char *symbol)
1717{
1718        struct module *owner;
1719        const struct kernel_symbol *sym;
1720
1721        preempt_disable();
1722        sym = find_symbol(symbol, &owner, NULL, true, true);
1723        if (sym && strong_try_module_get(owner))
1724                sym = NULL;
1725        preempt_enable();
1726
1727        return sym ? (void *)sym->value : NULL;
1728}
1729EXPORT_SYMBOL_GPL(__symbol_get);
1730
1731/*
1732 * Ensure that an exported symbol [global namespace] does not already exist
1733 * in the kernel or in some other module's exported symbol table.
1734 *
1735 * You must hold the module_mutex.
1736 */
1737static int verify_export_symbols(struct module *mod)
1738{
1739        unsigned int i;
1740        struct module *owner;
1741        const struct kernel_symbol *s;
1742        struct {
1743                const struct kernel_symbol *sym;
1744                unsigned int num;
1745        } arr[] = {
1746                { mod->syms, mod->num_syms },
1747                { mod->gpl_syms, mod->num_gpl_syms },
1748                { mod->gpl_future_syms, mod->num_gpl_future_syms },
1749#ifdef CONFIG_UNUSED_SYMBOLS
1750                { mod->unused_syms, mod->num_unused_syms },
1751                { mod->unused_gpl_syms, mod->num_unused_gpl_syms },
1752#endif
1753        };
1754
1755        for (i = 0; i < ARRAY_SIZE(arr); i++) {
1756                for (s = arr[i].sym; s < arr[i].sym + arr[i].num; s++) {
1757                        if (find_symbol(s->name, &owner, NULL, true, false)) {
1758                                printk(KERN_ERR
1759                                       "%s: exports duplicate symbol %s"
1760                                       " (owned by %s)\n",
1761                                       mod->name, s->name, module_name(owner));
1762                                return -ENOEXEC;
1763                        }
1764                }
1765        }
1766        return 0;
1767}
1768
1769/* Change all symbols so that st_value encodes the pointer directly. */
1770static int simplify_symbols(struct module *mod, const struct load_info *info)
1771{
1772        Elf_Shdr *symsec = &info->sechdrs[info->index.sym];
1773        Elf_Sym *sym = (void *)symsec->sh_addr;
1774        unsigned long secbase;
1775        unsigned int i;
1776        int ret = 0;
1777        const struct kernel_symbol *ksym;
1778
1779        for (i = 1; i < symsec->sh_size / sizeof(Elf_Sym); i++) {
1780                const char *name = info->strtab + sym[i].st_name;
1781
1782                switch (sym[i].st_shndx) {
1783                case SHN_COMMON:
1784                        /* We compiled with -fno-common.  These are not
1785                           supposed to happen.  */
1786                        DEBUGP("Common symbol: %s\n", name);
1787                        printk("%s: please compile with -fno-common\n",
1788                               mod->name);
1789                        ret = -ENOEXEC;
1790                        break;
1791
1792                case SHN_ABS:
1793                        /* Don't need to do anything */
1794                        DEBUGP("Absolute symbol: 0x%08lx\n",
1795                               (long)sym[i].st_value);
1796                        break;
1797
1798                case SHN_UNDEF:
1799                        ksym = resolve_symbol_wait(mod, info, name);
1800                        /* Ok if resolved.  */
1801                        if (ksym && !IS_ERR(ksym)) {
1802                                sym[i].st_value = ksym->value;
1803                                break;
1804                        }
1805
1806                        /* Ok if weak.  */
1807                        if (!ksym && ELF_ST_BIND(sym[i].st_info) == STB_WEAK)
1808                                break;
1809
1810                        printk(KERN_WARNING "%s: Unknown symbol %s (err %li)\n",
1811                               mod->name, name, PTR_ERR(ksym));
1812                        ret = PTR_ERR(ksym) ?: -ENOENT;
1813                        break;
1814
1815                default:
1816                        /* Divert to percpu allocation if a percpu var. */
1817                        if (sym[i].st_shndx == info->index.pcpu)
1818                                secbase = (unsigned long)mod_percpu(mod);
1819                        else
1820                                secbase = info->sechdrs[sym[i].st_shndx].sh_addr;
1821                        sym[i].st_value += secbase;
1822                        break;
1823                }
1824        }
1825
1826        return ret;
1827}
1828
1829static int apply_relocations(struct module *mod, const struct load_info *info)
1830{
1831        unsigned int i;
1832        int err = 0;
1833
1834        /* Now do relocations. */
1835        for (i = 1; i < info->hdr->e_shnum; i++) {
1836                unsigned int infosec = info->sechdrs[i].sh_info;
1837
1838                /* Not a valid relocation section? */
1839                if (infosec >= info->hdr->e_shnum)
1840                        continue;
1841
1842                /* Don't bother with non-allocated sections */
1843                if (!(info->sechdrs[infosec].sh_flags & SHF_ALLOC))
1844                        continue;
1845
1846                if (info->sechdrs[i].sh_type == SHT_REL)
1847                        err = apply_relocate(info->sechdrs, info->strtab,
1848                                             info->index.sym, i, mod);
1849                else if (info->sechdrs[i].sh_type == SHT_RELA)
1850                        err = apply_relocate_add(info->sechdrs, info->strtab,
1851                                                 info->index.sym, i, mod);
1852                if (err < 0)
1853                        break;
1854        }
1855        return err;
1856}
1857
1858/* Additional bytes needed by arch in front of individual sections */
1859unsigned int __weak arch_mod_section_prepend(struct module *mod,
1860                                             unsigned int section)
1861{
1862        /* default implementation just returns zero */
1863        return 0;
1864}
1865
1866/* Update size with this section: return offset. */
1867static long get_offset(struct module *mod, unsigned int *size,
1868                       Elf_Shdr *sechdr, unsigned int section)
1869{
1870        long ret;
1871
1872        *size += arch_mod_section_prepend(mod, section);
1873        ret = ALIGN(*size, sechdr->sh_addralign ?: 1);
1874        *size = ret + sechdr->sh_size;
1875        return ret;
1876}
1877
1878/* Lay out the SHF_ALLOC sections in a way not dissimilar to how ld
1879   might -- code, read-only data, read-write data, small data.  Tally
1880   sizes, and place the offsets into sh_entsize fields: high bit means it
1881   belongs in init. */
1882static void layout_sections(struct module *mod, struct load_info *info)
1883{
1884        static unsigned long const masks[][2] = {
1885                /* NOTE: all executable code must be the first section
1886                 * in this array; otherwise modify the text_size
1887                 * finder in the two loops below */
1888                { SHF_EXECINSTR | SHF_ALLOC, ARCH_SHF_SMALL },
1889                { SHF_ALLOC, SHF_WRITE | ARCH_SHF_SMALL },
1890                { SHF_WRITE | SHF_ALLOC, ARCH_SHF_SMALL },
1891                { ARCH_SHF_SMALL | SHF_ALLOC, 0 }
1892        };
1893        unsigned int m, i;
1894
1895        for (i = 0; i < info->hdr->e_shnum; i++)
1896                info->sechdrs[i].sh_entsize = ~0UL;
1897
1898        DEBUGP("Core section allocation order:\n");
1899        for (m = 0; m < ARRAY_SIZE(masks); ++m) {
1900                for (i = 0; i < info->hdr->e_shnum; ++i) {
1901                        Elf_Shdr *s = &info->sechdrs[i];
1902                        const char *sname = info->secstrings + s->sh_name;
1903
1904                        if ((s->sh_flags & masks[m][0]) != masks[m][0]
1905                            || (s->sh_flags & masks[m][1])
1906                            || s->sh_entsize != ~0UL
1907                            || strstarts(sname, ".init"))
1908                                continue;
1909                        s->sh_entsize = get_offset(mod, &mod->core_size, s, i);
1910                        DEBUGP("\t%s\n", name);
1911                }
1912                switch (m) {
1913                case 0: /* executable */
1914                        mod->core_size = debug_align(mod->core_size);
1915                        mod->core_text_size = mod->core_size;
1916                        break;
1917                case 1: /* RO: text and ro-data */
1918                        mod->core_size = debug_align(mod->core_size);
1919                        mod->core_ro_size = mod->core_size;
1920                        break;
1921                case 3: /* whole core */
1922                        mod->core_size = debug_align(mod->core_size);
1923                        break;
1924                }
1925        }
1926
1927        DEBUGP("Init section allocation order:\n");
1928        for (m = 0; m < ARRAY_SIZE(masks); ++m) {
1929                for (i = 0; i < info->hdr->e_shnum; ++i) {
1930                        Elf_Shdr *s = &info->sechdrs[i];
1931                        const char *sname = info->secstrings + s->sh_name;
1932
1933                        if ((s->sh_flags & masks[m][0]) != masks[m][0]
1934                            || (s->sh_flags & masks[m][1])
1935                            || s->sh_entsize != ~0UL
1936                            || !strstarts(sname, ".init"))
1937                                continue;
1938                        s->sh_entsize = (get_offset(mod, &mod->init_size, s, i)
1939                                         | INIT_OFFSET_MASK);
1940                        DEBUGP("\t%s\n", sname);
1941                }
1942                switch (m) {
1943                case 0: /* executable */
1944                        mod->init_size = debug_align(mod->init_size);
1945                        mod->init_text_size = mod->init_size;
1946                        break;
1947                case 1: /* RO: text and ro-data */
1948                        mod->init_size = debug_align(mod->init_size);
1949                        mod->init_ro_size = mod->init_size;
1950                        break;
1951                case 3: /* whole init */
1952                        mod->init_size = debug_align(mod->init_size);
1953                        break;
1954                }
1955        }
1956}
1957
1958static void set_license(struct module *mod, const char *license)
1959{
1960        if (!license)
1961                license = "unspecified";
1962
1963        if (!license_is_gpl_compatible(license)) {
1964                if (!test_taint(TAINT_PROPRIETARY_MODULE))
1965                        printk(KERN_WARNING "%s: module license '%s' taints "
1966                                "kernel.\n", mod->name, license);
1967                add_taint_module(mod, TAINT_PROPRIETARY_MODULE);
1968        }
1969}
1970
1971/* Parse tag=value strings from .modinfo section */
1972static char *next_string(char *string, unsigned long *secsize)
1973{
1974        /* Skip non-zero chars */
1975        while (string[0]) {
1976                string++;
1977                if ((*secsize)-- <= 1)
1978                        return NULL;
1979        }
1980
1981        /* Skip any zero padding. */
1982        while (!string[0]) {
1983                string++;
1984                if ((*secsize)-- <= 1)
1985                        return NULL;
1986        }
1987        return string;
1988}
1989
1990static char *get_modinfo(struct load_info *info, const char *tag)
1991{
1992        char *p;
1993        unsigned int taglen = strlen(tag);
1994        Elf_Shdr *infosec = &info->sechdrs[info->index.info];
1995        unsigned long size = infosec->sh_size;
1996
1997        for (p = (char *)infosec->sh_addr; p; p = next_string(p, &size)) {
1998                if (strncmp(p, tag, taglen) == 0 && p[taglen] == '=')
1999                        return p + taglen + 1;
2000        }
2001        return NULL;
2002}
2003
2004static void setup_modinfo(struct module *mod, struct load_info *info)
2005{
2006        struct module_attribute *attr;
2007        int i;
2008
2009        for (i = 0; (attr = modinfo_attrs[i]); i++) {
2010                if (attr->setup)
2011                        attr->setup(mod, get_modinfo(info, attr->attr.name));
2012        }
2013}
2014
2015static void free_modinfo(struct module *mod)
2016{
2017        struct module_attribute *attr;
2018        int i;
2019
2020        for (i = 0; (attr = modinfo_attrs[i]); i++) {
2021                if (attr->free)
2022                        attr->free(mod);
2023        }
2024}
2025
2026#ifdef CONFIG_KALLSYMS
2027
2028/* lookup symbol in given range of kernel_symbols */
2029static const struct kernel_symbol *lookup_symbol(const char *name,
2030        const struct kernel_symbol *start,
2031        const struct kernel_symbol *stop)
2032{
2033        const struct kernel_symbol *ks = start;
2034        for (; ks < stop; ks++)
2035                if (strcmp(ks->name, name) == 0)
2036                        return ks;
2037        return NULL;
2038}
2039
2040static int is_exported(const char *name, unsigned long value,
2041                       const struct module *mod)
2042{
2043        const struct kernel_symbol *ks;
2044        if (!mod)
2045                ks = lookup_symbol(name, __start___ksymtab, __stop___ksymtab);
2046        else
2047                ks = lookup_symbol(name, mod->syms, mod->syms + mod->num_syms);
2048        return ks != NULL && ks->value == value;
2049}
2050
2051/* As per nm */
2052static char elf_type(const Elf_Sym *sym, const struct load_info *info)
2053{
2054        const Elf_Shdr *sechdrs = info->sechdrs;
2055
2056        if (ELF_ST_BIND(sym->st_info) == STB_WEAK) {
2057                if (ELF_ST_TYPE(sym->st_info) == STT_OBJECT)
2058                        return 'v';
2059                else
2060                        return 'w';
2061        }
2062        if (sym->st_shndx == SHN_UNDEF)
2063                return 'U';
2064        if (sym->st_shndx == SHN_ABS)
2065                return 'a';
2066        if (sym->st_shndx >= SHN_LORESERVE)
2067                return '?';
2068        if (sechdrs[sym->st_shndx].sh_flags & SHF_EXECINSTR)
2069                return 't';
2070        if (sechdrs[sym->st_shndx].sh_flags & SHF_ALLOC
2071            && sechdrs[sym->st_shndx].sh_type != SHT_NOBITS) {
2072                if (!(sechdrs[sym->st_shndx].sh_flags & SHF_WRITE))
2073                        return 'r';
2074                else if (sechdrs[sym->st_shndx].sh_flags & ARCH_SHF_SMALL)
2075                        return 'g';
2076                else
2077                        return 'd';
2078        }
2079        if (sechdrs[sym->st_shndx].sh_type == SHT_NOBITS) {
2080                if (sechdrs[sym->st_shndx].sh_flags & ARCH_SHF_SMALL)
2081                        return 's';
2082                else
2083                        return 'b';
2084        }
2085        if (strstarts(info->secstrings + sechdrs[sym->st_shndx].sh_name,
2086                      ".debug")) {
2087                return 'n';
2088        }
2089        return '?';
2090}
2091
2092static bool is_core_symbol(const Elf_Sym *src, const Elf_Shdr *sechdrs,
2093                           unsigned int shnum)
2094{
2095        const Elf_Shdr *sec;
2096
2097        if (src->st_shndx == SHN_UNDEF
2098            || src->st_shndx >= shnum
2099            || !src->st_name)
2100                return false;
2101
2102        sec = sechdrs + src->st_shndx;
2103        if (!(sec->sh_flags & SHF_ALLOC)
2104#ifndef CONFIG_KALLSYMS_ALL
2105            || !(sec->sh_flags & SHF_EXECINSTR)
2106#endif
2107            || (sec->sh_entsize & INIT_OFFSET_MASK))
2108                return false;
2109
2110        return true;
2111}
2112
2113static void layout_symtab(struct module *mod, struct load_info *info)
2114{
2115        Elf_Shdr *symsect = info->sechdrs + info->index.sym;
2116        Elf_Shdr *strsect = info->sechdrs + info->index.str;
2117        const Elf_Sym *src;
2118        unsigned int i, nsrc, ndst;
2119
2120        /* Put symbol section at end of init part of module. */
2121        symsect->sh_flags |= SHF_ALLOC;
2122        symsect->sh_entsize = get_offset(mod, &mod->init_size, symsect,
2123                                         info->index.sym) | INIT_OFFSET_MASK;
2124        DEBUGP("\t%s\n", info->secstrings + symsect->sh_name);
2125
2126        src = (void *)info->hdr + symsect->sh_offset;
2127        nsrc = symsect->sh_size / sizeof(*src);
2128        for (ndst = i = 1; i < nsrc; ++i, ++src)
2129                if (is_core_symbol(src, info->sechdrs, info->hdr->e_shnum)) {
2130                        unsigned int j = src->st_name;
2131
2132                        while (!__test_and_set_bit(j, info->strmap)
2133                               && info->strtab[j])
2134                                ++j;
2135                        ++ndst;
2136                }
2137
2138        /* Append room for core symbols at end of core part. */
2139        info->symoffs = ALIGN(mod->core_size, symsect->sh_addralign ?: 1);
2140        mod->core_size = info->symoffs + ndst * sizeof(Elf_Sym);
2141
2142        /* Put string table section at end of init part of module. */
2143        strsect->sh_flags |= SHF_ALLOC;
2144        strsect->sh_entsize = get_offset(mod, &mod->init_size, strsect,
2145                                         info->index.str) | INIT_OFFSET_MASK;
2146        DEBUGP("\t%s\n", info->secstrings + strsect->sh_name);
2147
2148        /* Append room for core symbols' strings at end of core part. */
2149        info->stroffs = mod->core_size;
2150        __set_bit(0, info->strmap);
2151        mod->core_size += bitmap_weight(info->strmap, strsect->sh_size);
2152}
2153
2154static void add_kallsyms(struct module *mod, const struct load_info *info)
2155{
2156        unsigned int i, ndst;
2157        const Elf_Sym *src;
2158        Elf_Sym *dst;
2159        char *s;
2160        Elf_Shdr *symsec = &info->sechdrs[info->index.sym];
2161
2162        mod->symtab = (void *)symsec->sh_addr;
2163        mod->num_symtab = symsec->sh_size / sizeof(Elf_Sym);
2164        /* Make sure we get permanent strtab: don't use info->strtab. */
2165        mod->strtab = (void *)info->sechdrs[info->index.str].sh_addr;
2166
2167        /* Set types up while we still have access to sections. */
2168        for (i = 0; i < mod->num_symtab; i++)
2169                mod->symtab[i].st_info = elf_type(&mod->symtab[i], info);
2170
2171        mod->core_symtab = dst = mod->module_core + info->symoffs;
2172        src = mod->symtab;
2173        *dst = *src;
2174        for (ndst = i = 1; i < mod->num_symtab; ++i, ++src) {
2175                if (!is_core_symbol(src, info->sechdrs, info->hdr->e_shnum))
2176                        continue;
2177                dst[ndst] = *src;
2178                dst[ndst].st_name = bitmap_weight(info->strmap,
2179                                                  dst[ndst].st_name);
2180                ++ndst;
2181        }
2182        mod->core_num_syms = ndst;
2183
2184        mod->core_strtab = s = mod->module_core + info->stroffs;
2185        for (*s = 0, i = 1; i < info->sechdrs[info->index.str].sh_size; ++i)
2186                if (test_bit(i, info->strmap))
2187                        *++s = mod->strtab[i];
2188}
2189#else
2190static inline void layout_symtab(struct module *mod, struct load_info *info)
2191{
2192}
2193
2194static void add_kallsyms(struct module *mod, const struct load_info *info)
2195{
2196}
2197#endif /* CONFIG_KALLSYMS */
2198
2199static void dynamic_debug_setup(struct _ddebug *debug, unsigned int num)
2200{
2201        if (!debug)
2202                return;
2203#ifdef CONFIG_DYNAMIC_DEBUG
2204        if (ddebug_add_module(debug, num, debug->modname))
2205                printk(KERN_ERR "dynamic debug error adding module: %s\n",
2206                                        debug->modname);
2207#endif
2208}
2209
2210static void dynamic_debug_remove(struct _ddebug *debug)
2211{
2212        if (debug)
2213                ddebug_remove_module(debug->modname);
2214}
2215
2216static void *module_alloc_update_bounds(unsigned long size)
2217{
2218        void *ret = module_alloc(size);
2219
2220        if (ret) {
2221                mutex_lock(&module_mutex);
2222                /* Update module bounds. */
2223                if ((unsigned long)ret < module_addr_min)
2224                        module_addr_min = (unsigned long)ret;
2225                if ((unsigned long)ret + size > module_addr_max)
2226                        module_addr_max = (unsigned long)ret + size;
2227                mutex_unlock(&module_mutex);
2228        }
2229        return ret;
2230}
2231
2232#ifdef CONFIG_DEBUG_KMEMLEAK
2233static void kmemleak_load_module(const struct module *mod,
2234                                 const struct load_info *info)
2235{
2236        unsigned int i;
2237
2238        /* only scan the sections containing data */
2239        kmemleak_scan_area(mod, sizeof(struct module), GFP_KERNEL);
2240
2241        for (i = 1; i < info->hdr->e_shnum; i++) {
2242                const char *name = info->secstrings + info->sechdrs[i].sh_name;
2243                if (!(info->sechdrs[i].sh_flags & SHF_ALLOC))
2244                        continue;
2245                if (!strstarts(name, ".data") && !strstarts(name, ".bss"))
2246                        continue;
2247
2248                kmemleak_scan_area((void *)info->sechdrs[i].sh_addr,
2249                                   info->sechdrs[i].sh_size, GFP_KERNEL);
2250        }
2251}
2252#else
2253static inline void kmemleak_load_module(const struct module *mod,
2254                                        const struct load_info *info)
2255{
2256}
2257#endif
2258
2259/* Sets info->hdr and info->len. */
2260static int copy_and_check(struct load_info *info,
2261                          const void __user *umod, unsigned long len,
2262                          const char __user *uargs)
2263{
2264        int err;
2265        Elf_Ehdr *hdr;
2266
2267        if (len < sizeof(*hdr))
2268                return -ENOEXEC;
2269
2270        /* Suck in entire file: we'll want most of it. */
2271        /* vmalloc barfs on "unusual" numbers.  Check here */
2272        if (len > 64 * 1024 * 1024 || (hdr = vmalloc(len)) == NULL)
2273                return -ENOMEM;
2274
2275        if (copy_from_user(hdr, umod, len) != 0) {
2276                err = -EFAULT;
2277                goto free_hdr;
2278        }
2279
2280        /* Sanity checks against insmoding binaries or wrong arch,
2281           weird elf version */
2282        if (memcmp(hdr->e_ident, ELFMAG, SELFMAG) != 0
2283            || hdr->e_type != ET_REL
2284            || !elf_check_arch(hdr)
2285            || hdr->e_shentsize != sizeof(Elf_Shdr)) {
2286                err = -ENOEXEC;
2287                goto free_hdr;
2288        }
2289
2290        if (len < hdr->e_shoff + hdr->e_shnum * sizeof(Elf_Shdr)) {
2291                err = -ENOEXEC;
2292                goto free_hdr;
2293        }
2294
2295        info->hdr = hdr;
2296        info->len = len;
2297        return 0;
2298
2299free_hdr:
2300        vfree(hdr);
2301        return err;
2302}
2303
2304static void free_copy(struct load_info *info)
2305{
2306        vfree(info->hdr);
2307}
2308
2309static int rewrite_section_headers(struct load_info *info)
2310{
2311        unsigned int i;
2312
2313        /* This should always be true, but let's be sure. */
2314        info->sechdrs[0].sh_addr = 0;
2315
2316        for (i = 1; i < info->hdr->e_shnum; i++) {
2317                Elf_Shdr *shdr = &info->sechdrs[i];
2318                if (shdr->sh_type != SHT_NOBITS
2319                    && info->len < shdr->sh_offset + shdr->sh_size) {
2320                        printk(KERN_ERR "Module len %lu truncated\n",
2321                               info->len);
2322                        return -ENOEXEC;
2323                }
2324
2325                /* Mark all sections sh_addr with their address in the
2326                   temporary image. */
2327                shdr->sh_addr = (size_t)info->hdr + shdr->sh_offset;
2328
2329#ifndef CONFIG_MODULE_UNLOAD
2330                /* Don't load .exit sections */
2331                if (strstarts(info->secstrings+shdr->sh_name, ".exit"))
2332                        shdr->sh_flags &= ~(unsigned long)SHF_ALLOC;
2333#endif
2334        }
2335
2336        /* Track but don't keep modinfo and version sections. */
2337        info->index.vers = find_sec(info, "__versions");
2338        info->index.info = find_sec(info, ".modinfo");
2339        info->sechdrs[info->index.info].sh_flags &= ~(unsigned long)SHF_ALLOC;
2340        info->sechdrs[info->index.vers].sh_flags &= ~(unsigned long)SHF_ALLOC;
2341        return 0;
2342}
2343
2344/*
2345 * Set up our basic convenience variables (pointers to section headers,
2346 * search for module section index etc), and do some basic section
2347 * verification.
2348 *
2349 * Return the temporary module pointer (we'll replace it with the final
2350 * one when we move the module sections around).
2351 */
2352static struct module *setup_load_info(struct load_info *info)
2353{
2354        unsigned int i;
2355        int err;
2356        struct module *mod;
2357
2358        /* Set up the convenience variables */
2359        info->sechdrs = (void *)info->hdr + info->hdr->e_shoff;
2360        info->secstrings = (void *)info->hdr
2361                + info->sechdrs[info->hdr->e_shstrndx].sh_offset;
2362
2363        err = rewrite_section_headers(info);
2364        if (err)
2365                return ERR_PTR(err);
2366
2367        /* Find internal symbols and strings. */
2368        for (i = 1; i < info->hdr->e_shnum; i++) {
2369                if (info->sechdrs[i].sh_type == SHT_SYMTAB) {
2370                        info->index.sym = i;
2371                        info->index.str = info->sechdrs[i].sh_link;
2372                        info->strtab = (char *)info->hdr
2373                                + info->sechdrs[info->index.str].sh_offset;
2374                        break;
2375                }
2376        }
2377
2378        info->index.mod = find_sec(info, ".gnu.linkonce.this_module");
2379        if (!info->index.mod) {
2380                printk(KERN_WARNING "No module found in object\n");
2381                return ERR_PTR(-ENOEXEC);
2382        }
2383        /* This is temporary: point mod into copy of data. */
2384        mod = (void *)info->sechdrs[info->index.mod].sh_addr;
2385
2386        if (info->index.sym == 0) {
2387                printk(KERN_WARNING "%s: module has no symbols (stripped?)\n",
2388                       mod->name);
2389                return ERR_PTR(-ENOEXEC);
2390        }
2391
2392        info->index.pcpu = find_pcpusec(info);
2393
2394        /* Check module struct version now, before we try to use module. */
2395        if (!check_modstruct_version(info->sechdrs, info->index.vers, mod))
2396                return ERR_PTR(-ENOEXEC);
2397
2398        return mod;
2399}
2400
2401static int check_modinfo(struct module *mod, struct load_info *info)
2402{
2403        const char *modmagic = get_modinfo(info, "vermagic");
2404        int err;
2405
2406        /* This is allowed: modprobe --force will invalidate it. */
2407        if (!modmagic) {
2408                err = try_to_force_load(mod, "bad vermagic");
2409                if (err)
2410                        return err;
2411        } else if (!same_magic(modmagic, vermagic, info->index.vers)) {
2412                printk(KERN_ERR "%s: version magic '%s' should be '%s'\n",
2413                       mod->name, modmagic, vermagic);
2414                return -ENOEXEC;
2415        }
2416
2417        if (get_modinfo(info, "staging")) {
2418                add_taint_module(mod, TAINT_CRAP);
2419                printk(KERN_WARNING "%s: module is from the staging directory,"
2420                       " the quality is unknown, you have been warned.\n",
2421                       mod->name);
2422        }
2423
2424        /* Set up license info based on the info section */
2425        set_license(mod, get_modinfo(info, "license"));
2426
2427        return 0;
2428}
2429
2430static void find_module_sections(struct module *mod, struct load_info *info)
2431{
2432        mod->kp = section_objs(info, "__param",
2433                               sizeof(*mod->kp), &mod->num_kp);
2434        mod->syms = section_objs(info, "__ksymtab",
2435                                 sizeof(*mod->syms), &mod->num_syms);
2436        mod->crcs = section_addr(info, "__kcrctab");
2437        mod->gpl_syms = section_objs(info, "__ksymtab_gpl",
2438                                     sizeof(*mod->gpl_syms),
2439                                     &mod->num_gpl_syms);
2440        mod->gpl_crcs = section_addr(info, "__kcrctab_gpl");
2441        mod->gpl_future_syms = section_objs(info,
2442                                            "__ksymtab_gpl_future",
2443                                            sizeof(*mod->gpl_future_syms),
2444                                            &mod->num_gpl_future_syms);
2445        mod->gpl_future_crcs = section_addr(info, "__kcrctab_gpl_future");
2446
2447#ifdef CONFIG_UNUSED_SYMBOLS
2448        mod->unused_syms = section_objs(info, "__ksymtab_unused",
2449                                        sizeof(*mod->unused_syms),
2450                                        &mod->num_unused_syms);
2451        mod->unused_crcs = section_addr(info, "__kcrctab_unused");
2452        mod->unused_gpl_syms = section_objs(info, "__ksymtab_unused_gpl",
2453                                            sizeof(*mod->unused_gpl_syms),
2454                                            &mod->num_unused_gpl_syms);
2455        mod->unused_gpl_crcs = section_addr(info, "__kcrctab_unused_gpl");
2456#endif
2457#ifdef CONFIG_CONSTRUCTORS
2458        mod->ctors = section_objs(info, ".ctors",
2459                                  sizeof(*mod->ctors), &mod->num_ctors);
2460#endif
2461
2462#ifdef CONFIG_TRACEPOINTS
2463        mod->tracepoints_ptrs = section_objs(info, "__tracepoints_ptrs",
2464                                             sizeof(*mod->tracepoints_ptrs),
2465                                             &mod->num_tracepoints);
2466#endif
2467#ifdef HAVE_JUMP_LABEL
2468        mod->jump_entries = section_objs(info, "__jump_table",
2469                                        sizeof(*mod->jump_entries),
2470                                        &mod->num_jump_entries);
2471#endif
2472#ifdef CONFIG_EVENT_TRACING
2473        mod->trace_events = section_objs(info, "_ftrace_events",
2474                                         sizeof(*mod->trace_events),
2475                                         &mod->num_trace_events);
2476        /*
2477         * This section contains pointers to allocated objects in the trace
2478         * code and not scanning it leads to false positives.
2479         */
2480        kmemleak_scan_area(mod->trace_events, sizeof(*mod->trace_events) *
2481                           mod->num_trace_events, GFP_KERNEL);
2482#endif
2483#ifdef CONFIG_TRACING
2484        mod->trace_bprintk_fmt_start = section_objs(info, "__trace_printk_fmt",
2485                                         sizeof(*mod->trace_bprintk_fmt_start),
2486                                         &mod->num_trace_bprintk_fmt);
2487        /*
2488         * This section contains pointers to allocated objects in the trace
2489         * code and not scanning it leads to false positives.
2490         */
2491        kmemleak_scan_area(mod->trace_bprintk_fmt_start,
2492                           sizeof(*mod->trace_bprintk_fmt_start) *
2493                           mod->num_trace_bprintk_fmt, GFP_KERNEL);
2494#endif
2495#ifdef CONFIG_FTRACE_MCOUNT_RECORD
2496        /* sechdrs[0].sh_size is always zero */
2497        mod->ftrace_callsites = section_objs(info, "__mcount_loc",
2498                                             sizeof(*mod->ftrace_callsites),
2499                                             &mod->num_ftrace_callsites);
2500#endif
2501
2502        mod->extable = section_objs(info, "__ex_table",
2503                                    sizeof(*mod->extable), &mod->num_exentries);
2504
2505        if (section_addr(info, "__obsparm"))
2506                printk(KERN_WARNING "%s: Ignoring obsolete parameters\n",
2507                       mod->name);
2508
2509        info->debug = section_objs(info, "__verbose",
2510                                   sizeof(*info->debug), &info->num_debug);
2511}
2512
2513static int move_module(struct module *mod, struct load_info *info)
2514{
2515        int i;
2516        void *ptr;
2517
2518        /* Do the allocs. */
2519        ptr = module_alloc_update_bounds(mod->core_size);
2520        /*
2521         * The pointer to this block is stored in the module structure
2522         * which is inside the block. Just mark it as not being a
2523         * leak.
2524         */
2525        kmemleak_not_leak(ptr);
2526        if (!ptr)
2527                return -ENOMEM;
2528
2529        memset(ptr, 0, mod->core_size);
2530        mod->module_core = ptr;
2531
2532        ptr = module_alloc_update_bounds(mod->init_size);
2533        /*
2534         * The pointer to this block is stored in the module structure
2535         * which is inside the block. This block doesn't need to be
2536         * scanned as it contains data and code that will be freed
2537         * after the module is initialized.
2538         */
2539        kmemleak_ignore(ptr);
2540        if (!ptr && mod->init_size) {
2541                module_free(mod, mod->module_core);
2542                return -ENOMEM;
2543        }
2544        memset(ptr, 0, mod->init_size);
2545        mod->module_init = ptr;
2546
2547        /* Transfer each section which specifies SHF_ALLOC */
2548        DEBUGP("final section addresses:\n");
2549        for (i = 0; i < info->hdr->e_shnum; i++) {
2550                void *dest;
2551                Elf_Shdr *shdr = &info->sechdrs[i];
2552
2553                if (!(shdr->sh_flags & SHF_ALLOC))
2554                        continue;
2555
2556                if (shdr->sh_entsize & INIT_OFFSET_MASK)
2557                        dest = mod->module_init
2558                                + (shdr->sh_entsize & ~INIT_OFFSET_MASK);
2559                else
2560                        dest = mod->module_core + shdr->sh_entsize;
2561
2562                if (shdr->sh_type != SHT_NOBITS)
2563                        memcpy(dest, (void *)shdr->sh_addr, shdr->sh_size);
2564                /* Update sh_addr to point to copy in image. */
2565                shdr->sh_addr = (unsigned long)dest;
2566                DEBUGP("\t0x%lx %s\n",
2567                       shdr->sh_addr, info->secstrings + shdr->sh_name);
2568        }
2569
2570        return 0;
2571}
2572
2573static int check_module_license_and_versions(struct module *mod)
2574{
2575        /*
2576         * ndiswrapper is under GPL by itself, but loads proprietary modules.
2577         * Don't use add_taint_module(), as it would prevent ndiswrapper from
2578         * using GPL-only symbols it needs.
2579         */
2580        if (strcmp(mod->name, "ndiswrapper") == 0)
2581                add_taint(TAINT_PROPRIETARY_MODULE);
2582
2583        /* driverloader was caught wrongly pretending to be under GPL */
2584        if (strcmp(mod->name, "driverloader") == 0)
2585                add_taint_module(mod, TAINT_PROPRIETARY_MODULE);
2586
2587#ifdef CONFIG_MODVERSIONS
2588        if ((mod->num_syms && !mod->crcs)
2589            || (mod->num_gpl_syms && !mod->gpl_crcs)
2590            || (mod->num_gpl_future_syms && !mod->gpl_future_crcs)
2591#ifdef CONFIG_UNUSED_SYMBOLS
2592            || (mod->num_unused_syms && !mod->unused_crcs)
2593            || (mod->num_unused_gpl_syms && !mod->unused_gpl_crcs)
2594#endif
2595                ) {
2596                return try_to_force_load(mod,
2597                                         "no versions for exported symbols");
2598        }
2599#endif
2600        return 0;
2601}
2602
2603static void flush_module_icache(const struct module *mod)
2604{
2605        mm_segment_t old_fs;
2606
2607        /* flush the icache in correct context */
2608        old_fs = get_fs();
2609        set_fs(KERNEL_DS);
2610
2611        /*
2612         * Flush the instruction cache, since we've played with text.
2613         * Do it before processing of module parameters, so the module
2614         * can provide parameter accessor functions of its own.
2615         */
2616        if (mod->module_init)
2617                flush_icache_range((unsigned long)mod->module_init,
2618                                   (unsigned long)mod->module_init
2619                                   + mod->init_size);
2620        flush_icache_range((unsigned long)mod->module_core,
2621                           (unsigned long)mod->module_core + mod->core_size);
2622
2623        set_fs(old_fs);
2624}
2625
2626static struct module *layout_and_allocate(struct load_info *info)
2627{
2628        /* Module within temporary copy. */
2629        struct module *mod;
2630        Elf_Shdr *pcpusec;
2631        int err;
2632
2633        mod = setup_load_info(info);
2634        if (IS_ERR(mod))
2635                return mod;
2636
2637        err = check_modinfo(mod, info);
2638        if (err)
2639                return ERR_PTR(err);
2640
2641        /* Allow arches to frob section contents and sizes.  */
2642        err = module_frob_arch_sections(info->hdr, info->sechdrs,
2643                                        info->secstrings, mod);
2644        if (err < 0)
2645                goto out;
2646
2647        pcpusec = &info->sechdrs[info->index.pcpu];
2648        if (pcpusec->sh_size) {
2649                /* We have a special allocation for this section. */
2650                err = percpu_modalloc(mod,
2651                                      pcpusec->sh_size, pcpusec->sh_addralign);
2652                if (err)
2653                        goto out;
2654                pcpusec->sh_flags &= ~(unsigned long)SHF_ALLOC;
2655        }
2656
2657        /* Determine total sizes, and put offsets in sh_entsize.  For now
2658           this is done generically; there doesn't appear to be any
2659           special cases for the architectures. */
2660        layout_sections(mod, info);
2661
2662        info->strmap = kzalloc(BITS_TO_LONGS(info->sechdrs[info->index.str].sh_size)
2663                         * sizeof(long), GFP_KERNEL);
2664        if (!info->strmap) {
2665                err = -ENOMEM;
2666                goto free_percpu;
2667        }
2668        layout_symtab(mod, info);
2669
2670        /* Allocate and move to the final place */
2671        err = move_module(mod, info);
2672        if (err)
2673                goto free_strmap;
2674
2675        /* Module has been copied to its final place now: return it. */
2676        mod = (void *)info->sechdrs[info->index.mod].sh_addr;
2677        kmemleak_load_module(mod, info);
2678        return mod;
2679
2680free_strmap:
2681        kfree(info->strmap);
2682free_percpu:
2683        percpu_modfree(mod);
2684out:
2685        return ERR_PTR(err);
2686}
2687
2688/* mod is no longer valid after this! */
2689static void module_deallocate(struct module *mod, struct load_info *info)
2690{
2691        kfree(info->strmap);
2692        percpu_modfree(mod);
2693        module_free(mod, mod->module_init);
2694        module_free(mod, mod->module_core);
2695}
2696
2697static int post_relocation(struct module *mod, const struct load_info *info)
2698{
2699        /* Sort exception table now relocations are done. */
2700        sort_extable(mod->extable, mod->extable + mod->num_exentries);
2701
2702        /* Copy relocated percpu area over. */
2703        percpu_modcopy(mod, (void *)info->sechdrs[info->index.pcpu].sh_addr,
2704                       info->sechdrs[info->index.pcpu].sh_size);
2705
2706        /* Setup kallsyms-specific fields. */
2707        add_kallsyms(mod, info);
2708
2709        /* Arch-specific module finalizing. */
2710        return module_finalize(info->hdr, info->sechdrs, mod);
2711}
2712
2713/* Allocate and load the module: note that size of section 0 is always
2714   zero, and we rely on this for optional sections. */
2715static struct module *load_module(void __user *umod,
2716                                  unsigned long len,
2717                                  const char __user *uargs)
2718{
2719        struct load_info info = { NULL, };
2720        struct module *mod;
2721        long err;
2722
2723        DEBUGP("load_module: umod=%p, len=%lu, uargs=%p\n",
2724               umod, len, uargs);
2725
2726        /* Copy in the blobs from userspace, check they are vaguely sane. */
2727        err = copy_and_check(&info, umod, len, uargs);
2728        if (err)
2729                return ERR_PTR(err);
2730
2731        /* Figure out module layout, and allocate all the memory. */
2732        mod = layout_and_allocate(&info);
2733        if (IS_ERR(mod)) {
2734                err = PTR_ERR(mod);
2735                goto free_copy;
2736        }
2737
2738        /* Now module is in final location, initialize linked lists, etc. */
2739        err = module_unload_init(mod);
2740        if (err)
2741                goto free_module;
2742
2743        /* Now we've got everything in the final locations, we can
2744         * find optional sections. */
2745        find_module_sections(mod, &info);
2746
2747        err = check_module_license_and_versions(mod);
2748        if (err)
2749                goto free_unload;
2750
2751        /* Set up MODINFO_ATTR fields */
2752        setup_modinfo(mod, &info);
2753
2754        /* Fix up syms, so that st_value is a pointer to location. */
2755        err = simplify_symbols(mod, &info);
2756        if (err < 0)
2757                goto free_modinfo;
2758
2759        err = apply_relocations(mod, &info);
2760        if (err < 0)
2761                goto free_modinfo;
2762
2763        err = post_relocation(mod, &info);
2764        if (err < 0)
2765                goto free_modinfo;
2766
2767        flush_module_icache(mod);
2768
2769        /* Now copy in args */
2770        mod->args = strndup_user(uargs, ~0UL >> 1);
2771        if (IS_ERR(mod->args)) {
2772                err = PTR_ERR(mod->args);
2773                goto free_arch_cleanup;
2774        }
2775
2776        /* Mark state as coming so strong_try_module_get() ignores us. */
2777        mod->state = MODULE_STATE_COMING;
2778
2779        /* Now sew it into the lists so we can get lockdep and oops
2780         * info during argument parsing.  Noone should access us, since
2781         * strong_try_module_get() will fail.
2782         * lockdep/oops can run asynchronous, so use the RCU list insertion
2783         * function to insert in a way safe to concurrent readers.
2784         * The mutex protects against concurrent writers.
2785         */
2786        mutex_lock(&module_mutex);
2787        if (find_module(mod->name)) {
2788                err = -EEXIST;
2789                goto unlock;
2790        }
2791
2792        /* This has to be done once we're sure module name is unique. */
2793        if (!mod->taints)
2794                dynamic_debug_setup(info.debug, info.num_debug);
2795
2796        /* Find duplicate symbols */
2797        err = verify_export_symbols(mod);
2798        if (err < 0)
2799                goto ddebug;
2800
2801        module_bug_finalize(info.hdr, info.sechdrs, mod);
2802        list_add_rcu(&mod->list, &modules);
2803        mutex_unlock(&module_mutex);
2804
2805        /* Module is ready to execute: parsing args may do that. */
2806        err = parse_args(mod->name, mod->args, mod->kp, mod->num_kp, NULL);
2807        if (err < 0)
2808                goto unlink;
2809
2810        /* Link in to syfs. */
2811        err = mod_sysfs_setup(mod, &info, mod->kp, mod->num_kp);
2812        if (err < 0)
2813                goto unlink;
2814
2815        /* Get rid of temporary copy and strmap. */
2816        kfree(info.strmap);
2817        free_copy(&info);
2818
2819        /* Done! */
2820        trace_module_load(mod);
2821        return mod;
2822
2823 unlink:
2824        mutex_lock(&module_mutex);
2825        /* Unlink carefully: kallsyms could be walking list. */
2826        list_del_rcu(&mod->list);
2827        module_bug_cleanup(mod);
2828
2829 ddebug:
2830        if (!mod->taints)
2831                dynamic_debug_remove(info.debug);
2832 unlock:
2833        mutex_unlock(&module_mutex);
2834        synchronize_sched();
2835        kfree(mod->args);
2836 free_arch_cleanup:
2837        module_arch_cleanup(mod);
2838 free_modinfo:
2839        free_modinfo(mod);
2840 free_unload:
2841        module_unload_free(mod);
2842 free_module:
2843        module_deallocate(mod, &info);
2844 free_copy:
2845        free_copy(&info);
2846        return ERR_PTR(err);
2847}
2848
2849/* Call module constructors. */
2850static void do_mod_ctors(struct module *mod)
2851{
2852#ifdef CONFIG_CONSTRUCTORS
2853        unsigned long i;
2854
2855        for (i = 0; i < mod->num_ctors; i++)
2856                mod->ctors[i]();
2857#endif
2858}
2859
2860/* This is where the real work happens */
2861SYSCALL_DEFINE3(init_module, void __user *, umod,
2862                unsigned long, len, const char __user *, uargs)
2863{
2864        struct module *mod;
2865        int ret = 0;
2866
2867        /* Must have permission */
2868        if (!capable(CAP_SYS_MODULE) || modules_disabled)
2869                return -EPERM;
2870
2871        /* Do all the hard work */
2872        mod = load_module(umod, len, uargs);
2873        if (IS_ERR(mod))
2874                return PTR_ERR(mod);
2875
2876        blocking_notifier_call_chain(&module_notify_list,
2877                        MODULE_STATE_COMING, mod);
2878
2879        /* Set RO and NX regions for core */
2880        set_section_ro_nx(mod->module_core,
2881                                mod->core_text_size,
2882                                mod->core_ro_size,
2883                                mod->core_size);
2884
2885        /* Set RO and NX regions for init */
2886        set_section_ro_nx(mod->module_init,
2887                                mod->init_text_size,
2888                                mod->init_ro_size,
2889                                mod->init_size);
2890
2891        do_mod_ctors(mod);
2892        /* Start the module */
2893        if (mod->init != NULL)
2894                ret = do_one_initcall(mod->init);
2895        if (ret < 0) {
2896                /* Init routine failed: abort.  Try to protect us from
2897                   buggy refcounters. */
2898                mod->state = MODULE_STATE_GOING;
2899                synchronize_sched();
2900                module_put(mod);
2901                blocking_notifier_call_chain(&module_notify_list,
2902                                             MODULE_STATE_GOING, mod);
2903                free_module(mod);
2904                wake_up(&module_wq);
2905                return ret;
2906        }
2907        if (ret > 0) {
2908                printk(KERN_WARNING
2909"%s: '%s'->init suspiciously returned %d, it should follow 0/-E convention\n"
2910"%s: loading module anyway...\n",
2911                       __func__, mod->name, ret,
2912                       __func__);
2913                dump_stack();
2914        }
2915
2916        /* Now it's a first class citizen!  Wake up anyone waiting for it. */
2917        mod->state = MODULE_STATE_LIVE;
2918        wake_up(&module_wq);
2919        blocking_notifier_call_chain(&module_notify_list,
2920                                     MODULE_STATE_LIVE, mod);
2921
2922        /* We need to finish all async code before the module init sequence is done */
2923        async_synchronize_full();
2924
2925        mutex_lock(&module_mutex);
2926        /* Drop initial reference. */
2927        module_put(mod);
2928        trim_init_extable(mod);
2929#ifdef CONFIG_KALLSYMS
2930        mod->num_symtab = mod->core_num_syms;
2931        mod->symtab = mod->core_symtab;
2932        mod->strtab = mod->core_strtab;
2933#endif
2934        unset_section_ro_nx(mod, mod->module_init);
2935        module_free(mod, mod->module_init);
2936        mod->module_init = NULL;
2937        mod->init_size = 0;
2938        mod->init_text_size = 0;
2939        mutex_unlock(&module_mutex);
2940
2941        return 0;
2942}
2943
2944static inline int within(unsigned long addr, void *start, unsigned long size)
2945{
2946        return ((void *)addr >= start && (void *)addr < start + size);
2947}
2948
2949#ifdef CONFIG_KALLSYMS
2950/*
2951 * This ignores the intensely annoying "mapping symbols" found
2952 * in ARM ELF files: $a, $t and $d.
2953 */
2954static inline int is_arm_mapping_symbol(const char *str)
2955{
2956        return str[0] == '$' && strchr("atd", str[1])
2957               && (str[2] == '\0' || str[2] == '.');
2958}
2959
2960static const char *get_ksymbol(struct module *mod,
2961                               unsigned long addr,
2962                               unsigned long *size,
2963                               unsigned long *offset)
2964{
2965        unsigned int i, best = 0;
2966        unsigned long nextval;
2967
2968        /* At worse, next value is at end of module */
2969        if (within_module_init(addr, mod))
2970                nextval = (unsigned long)mod->module_init+mod->init_text_size;
2971        else
2972                nextval = (unsigned long)mod->module_core+mod->core_text_size;
2973
2974        /* Scan for closest preceeding symbol, and next symbol. (ELF
2975           starts real symbols at 1). */
2976        for (i = 1; i < mod->num_symtab; i++) {
2977                if (mod->symtab[i].st_shndx == SHN_UNDEF)
2978                        continue;
2979
2980                /* We ignore unnamed symbols: they're uninformative
2981                 * and inserted at a whim. */
2982                if (mod->symtab[i].st_value <= addr
2983                    && mod->symtab[i].st_value > mod->symtab[best].st_value
2984                    && *(mod->strtab + mod->symtab[i].st_name) != '\0'
2985                    && !is_arm_mapping_symbol(mod->strtab + mod->symtab[i].st_name))
2986                        best = i;
2987                if (mod->symtab[i].st_value > addr
2988                    && mod->symtab[i].st_value < nextval
2989                    && *(mod->strtab + mod->symtab[i].st_name) != '\0'
2990                    && !is_arm_mapping_symbol(mod->strtab + mod->symtab[i].st_name))
2991                        nextval = mod->symtab[i].st_value;
2992        }
2993
2994        if (!best)
2995                return NULL;
2996
2997        if (size)
2998                *size = nextval - mod->symtab[best].st_value;
2999        if (offset)
3000                *offset = addr - mod->symtab[best].st_value;
3001        return mod->strtab + mod->symtab[best].st_name;
3002}
3003
3004/* For kallsyms to ask for address resolution.  NULL means not found.  Careful
3005 * not to lock to avoid deadlock on oopses, simply disable preemption. */
3006const char *module_address_lookup(unsigned long addr,
3007                            unsigned long *size,
3008                            unsigned long *offset,
3009                            char **modname,
3010                            char *namebuf)
3011{
3012        struct module *mod;
3013        const char *ret = NULL;
3014
3015        preempt_disable();
3016        list_for_each_entry_rcu(mod, &modules, list) {
3017                if (within_module_init(addr, mod) ||
3018                    within_module_core(addr, mod)) {
3019                        if (modname)
3020                                *modname = mod->name;
3021                        ret = get_ksymbol(mod, addr, size, offset);
3022                        break;
3023                }
3024        }
3025        /* Make a copy in here where it's safe */
3026        if (ret) {
3027                strncpy(namebuf, ret, KSYM_NAME_LEN - 1);
3028                ret = namebuf;
3029        }
3030        preempt_enable();
3031        return ret;
3032}
3033
3034int lookup_module_symbol_name(unsigned long addr, char *symname)
3035{
3036        struct module *mod;
3037
3038        preempt_disable();
3039        list_for_each_entry_rcu(mod, &modules, list) {
3040                if (within_module_init(addr, mod) ||
3041                    within_module_core(addr, mod)) {
3042                        const char *sym;
3043
3044                        sym = get_ksymbol(mod, addr, NULL, NULL);
3045                        if (!sym)
3046                                goto out;
3047                        strlcpy(symname, sym, KSYM_NAME_LEN);
3048                        preempt_enable();
3049                        return 0;
3050                }
3051        }
3052out:
3053        preempt_enable();
3054        return -ERANGE;
3055}
3056
3057int lookup_module_symbol_attrs(unsigned long addr, unsigned long *size,
3058                        unsigned long *offset, char *modname, char *name)
3059{
3060        struct module *mod;
3061
3062        preempt_disable();
3063        list_for_each_entry_rcu(mod, &modules, list) {
3064                if (within_module_init(addr, mod) ||
3065                    within_module_core(addr, mod)) {
3066                        const char *sym;
3067
3068                        sym = get_ksymbol(mod, addr, size, offset);
3069                        if (!sym)
3070                                goto out;
3071                        if (modname)
3072                                strlcpy(modname, mod->name, MODULE_NAME_LEN);
3073                        if (name)
3074                                strlcpy(name, sym, KSYM_NAME_LEN);
3075                        preempt_enable();
3076                        return 0;
3077                }
3078        }
3079out:
3080        preempt_enable();
3081        return -ERANGE;
3082}
3083
3084int module_get_kallsym(unsigned int symnum, unsigned long *value, char *type,
3085                        char *name, char *module_name, int *exported)
3086{
3087        struct module *mod;
3088
3089        preempt_disable();
3090        list_for_each_entry_rcu(mod, &modules, list) {
3091                if (symnum < mod->num_symtab) {
3092                        *value = mod->symtab[symnum].st_value;
3093                        *type = mod->symtab[symnum].st_info;
3094                        strlcpy(name, mod->strtab + mod->symtab[symnum].st_name,
3095                                KSYM_NAME_LEN);
3096                        strlcpy(module_name, mod->name, MODULE_NAME_LEN);
3097                        *exported = is_exported(name, *value, mod);
3098                        preempt_enable();
3099                        return 0;
3100                }
3101                symnum -= mod->num_symtab;
3102        }
3103        preempt_enable();
3104        return -ERANGE;
3105}
3106
3107static unsigned long mod_find_symname(struct module *mod, const char *name)
3108{
3109        unsigned int i;
3110
3111        for (i = 0; i < mod->num_symtab; i++)
3112                if (strcmp(name, mod->strtab+mod->symtab[i].st_name) == 0 &&
3113                    mod->symtab[i].st_info != 'U')
3114                        return mod->symtab[i].st_value;
3115        return 0;
3116}
3117
3118/* Look for this name: can be of form module:name. */
3119unsigned long module_kallsyms_lookup_name(const char *name)
3120{
3121        struct module *mod;
3122        char *colon;
3123        unsigned long ret = 0;
3124
3125        /* Don't lock: we're in enough trouble already. */
3126        preempt_disable();
3127        if ((colon = strchr(name, ':')) != NULL) {
3128                *colon = '\0';
3129                if ((mod = find_module(name)) != NULL)
3130                        ret = mod_find_symname(mod, colon+1);
3131                *colon = ':';
3132        } else {
3133                list_for_each_entry_rcu(mod, &modules, list)
3134                        if ((ret = mod_find_symname(mod, name)) != 0)
3135                                break;
3136        }
3137        preempt_enable();
3138        return ret;
3139}
3140
3141int module_kallsyms_on_each_symbol(int (*fn)(void *, const char *,
3142                                             struct module *, unsigned long),
3143                                   void *data)
3144{
3145        struct module *mod;
3146        unsigned int i;
3147        int ret;
3148
3149        list_for_each_entry(mod, &modules, list) {
3150                for (i = 0; i < mod->num_symtab; i++) {
3151                        ret = fn(data, mod->strtab + mod->symtab[i].st_name,
3152                                 mod, mod->symtab[i].st_value);
3153                        if (ret != 0)
3154                                return ret;
3155                }
3156        }
3157        return 0;
3158}
3159#endif /* CONFIG_KALLSYMS */
3160
3161static char *module_flags(struct module *mod, char *buf)
3162{
3163        int bx = 0;
3164
3165        if (mod->taints ||
3166            mod->state == MODULE_STATE_GOING ||
3167            mod->state == MODULE_STATE_COMING) {
3168                buf[bx++] = '(';
3169                if (mod->taints & (1 << TAINT_PROPRIETARY_MODULE))
3170                        buf[bx++] = 'P';
3171                if (mod->taints & (1 << TAINT_FORCED_MODULE))
3172                        buf[bx++] = 'F';
3173                if (mod->taints & (1 << TAINT_CRAP))
3174                        buf[bx++] = 'C';
3175                /*
3176                 * TAINT_FORCED_RMMOD: could be added.
3177                 * TAINT_UNSAFE_SMP, TAINT_MACHINE_CHECK, TAINT_BAD_PAGE don't
3178                 * apply to modules.
3179                 */
3180
3181                /* Show a - for module-is-being-unloaded */
3182                if (mod->state == MODULE_STATE_GOING)
3183                        buf[bx++] = '-';
3184                /* Show a + for module-is-being-loaded */
3185                if (mod->state == MODULE_STATE_COMING)
3186                        buf[bx++] = '+';
3187                buf[bx++] = ')';
3188        }
3189        buf[bx] = '\0';
3190
3191        return buf;
3192}
3193
3194#ifdef CONFIG_PROC_FS
3195/* Called by the /proc file system to return a list of modules. */
3196static void *m_start(struct seq_file *m, loff_t *pos)
3197{
3198        mutex_lock(&module_mutex);
3199        return seq_list_start(&modules, *pos);
3200}
3201
3202static void *m_next(struct seq_file *m, void *p, loff_t *pos)
3203{
3204        return seq_list_next(p, &modules, pos);
3205}
3206
3207static void m_stop(struct seq_file *m, void *p)
3208{
3209        mutex_unlock(&module_mutex);
3210}
3211
3212static int m_show(struct seq_file *m, void *p)
3213{
3214        struct module *mod = list_entry(p, struct module, list);
3215        char buf[8];
3216
3217        seq_printf(m, "%s %u",
3218                   mod->name, mod->init_size + mod->core_size);
3219        print_unload_info(m, mod);
3220
3221        /* Informative for users. */
3222        seq_printf(m, " %s",
3223                   mod->state == MODULE_STATE_GOING ? "Unloading":
3224                   mod->state == MODULE_STATE_COMING ? "Loading":
3225                   "Live");
3226        /* Used by oprofile and other similar tools. */
3227        seq_printf(m, " 0x%p", mod->module_core);
3228
3229        /* Taints info */
3230        if (mod->taints)
3231                seq_printf(m, " %s", module_flags(mod, buf));
3232
3233        seq_printf(m, "\n");
3234        return 0;
3235}
3236
3237/* Format: modulename size refcount deps address
3238
3239   Where refcount is a number or -, and deps is a comma-separated list
3240   of depends or -.
3241*/
3242static const struct seq_operations modules_op = {
3243        .start  = m_start,
3244        .next   = m_next,
3245        .stop   = m_stop,
3246        .show   = m_show
3247};
3248
3249static int modules_open(struct inode *inode, struct file *file)
3250{
3251        return seq_open(file, &modules_op);
3252}
3253
3254static const struct file_operations proc_modules_operations = {
3255        .open           = modules_open,
3256        .read           = seq_read,
3257        .llseek         = seq_lseek,
3258        .release        = seq_release,
3259};
3260
3261static int __init proc_modules_init(void)
3262{
3263        proc_create("modules", 0, NULL, &proc_modules_operations);
3264        return 0;
3265}
3266module_init(proc_modules_init);
3267#endif
3268
3269/* Given an address, look for it in the module exception tables. */
3270const struct exception_table_entry *search_module_extables(unsigned long addr)
3271{
3272        const struct exception_table_entry *e = NULL;
3273        struct module *mod;
3274
3275        preempt_disable();
3276        list_for_each_entry_rcu(mod, &modules, list) {
3277                if (mod->num_exentries == 0)
3278                        continue;
3279
3280                e = search_extable(mod->extable,
3281                                   mod->extable + mod->num_exentries - 1,
3282                                   addr);
3283                if (e)
3284                        break;
3285        }
3286        preempt_enable();
3287
3288        /* Now, if we found one, we are running inside it now, hence
3289           we cannot unload the module, hence no refcnt needed. */
3290        return e;
3291}
3292
3293/*
3294 * is_module_address - is this address inside a module?
3295 * @addr: the address to check.
3296 *
3297 * See is_module_text_address() if you simply want to see if the address
3298 * is code (not data).
3299 */
3300bool is_module_address(unsigned long addr)
3301{
3302        bool ret;
3303
3304        preempt_disable();
3305        ret = __module_address(addr) != NULL;
3306        preempt_enable();
3307
3308        return ret;
3309}
3310
3311/*
3312 * __module_address - get the module which contains an address.
3313 * @addr: the address.
3314 *
3315 * Must be called with preempt disabled or module mutex held so that
3316 * module doesn't get freed during this.
3317 */
3318struct module *__module_address(unsigned long addr)
3319{
3320        struct module *mod;
3321
3322        if (addr < module_addr_min || addr > module_addr_max)
3323                return NULL;
3324
3325        list_for_each_entry_rcu(mod, &modules, list)
3326                if (within_module_core(addr, mod)
3327                    || within_module_init(addr, mod))
3328                        return mod;
3329        return NULL;
3330}
3331EXPORT_SYMBOL_GPL(__module_address);
3332
3333/*
3334 * is_module_text_address - is this address inside module code?
3335 * @addr: the address to check.
3336 *
3337 * See is_module_address() if you simply want to see if the address is
3338 * anywhere in a module.  See kernel_text_address() for testing if an
3339 * address corresponds to kernel or module code.
3340 */
3341bool is_module_text_address(unsigned long addr)
3342{
3343        bool ret;
3344
3345        preempt_disable();
3346        ret = __module_text_address(addr) != NULL;
3347        preempt_enable();
3348
3349        return ret;
3350}
3351
3352/*
3353 * __module_text_address - get the module whose code contains an address.
3354 * @addr: the address.
3355 *
3356 * Must be called with preempt disabled or module mutex held so that
3357 * module doesn't get freed during this.
3358 */
3359struct module *__module_text_address(unsigned long addr)
3360{
3361        struct module *mod = __module_address(addr);
3362        if (mod) {
3363                /* Make sure it's within the text section. */
3364                if (!within(addr, mod->module_init, mod->init_text_size)
3365                    && !within(addr, mod->module_core, mod->core_text_size))
3366                        mod = NULL;
3367        }
3368        return mod;
3369}
3370EXPORT_SYMBOL_GPL(__module_text_address);
3371
3372/* Don't grab lock, we're oopsing. */
3373void print_modules(void)
3374{
3375        struct module *mod;
3376        char buf[8];
3377
3378        printk(KERN_DEFAULT "Modules linked in:");
3379        /* Most callers should already have preempt disabled, but make sure */
3380        preempt_disable();
3381        list_for_each_entry_rcu(mod, &modules, list)
3382                printk(" %s%s", mod->name, module_flags(mod, buf));
3383        preempt_enable();
3384        if (last_unloaded_module[0])
3385                printk(" [last unloaded: %s]", last_unloaded_module);
3386        printk("\n");
3387}
3388
3389#ifdef CONFIG_MODVERSIONS
3390/* Generate the signature for all relevant module structures here.
3391 * If these change, we don't want to try to parse the module. */
3392void module_layout(struct module *mod,
3393                   struct modversion_info *ver,
3394                   struct kernel_param *kp,
3395                   struct kernel_symbol *ks,
3396                   struct tracepoint * const *tp)
3397{
3398}
3399EXPORT_SYMBOL(module_layout);
3400#endif
3401
3402#ifdef CONFIG_TRACEPOINTS
3403void module_update_tracepoints(void)
3404{
3405        struct module *mod;
3406
3407        mutex_lock(&module_mutex);
3408        list_for_each_entry(mod, &modules, list)
3409                if (!mod->taints)
3410                        tracepoint_update_probe_range(mod->tracepoints_ptrs,
3411                                mod->tracepoints_ptrs + mod->num_tracepoints);
3412        mutex_unlock(&module_mutex);
3413}
3414
3415/*
3416 * Returns 0 if current not found.
3417 * Returns 1 if current found.
3418 */
3419int module_get_iter_tracepoints(struct tracepoint_iter *iter)
3420{
3421        struct module *iter_mod;
3422        int found = 0;
3423
3424        mutex_lock(&module_mutex);
3425        list_for_each_entry(iter_mod, &modules, list) {
3426                if (!iter_mod->taints) {
3427                        /*
3428                         * Sorted module list
3429                         */
3430                        if (iter_mod < iter->module)
3431                                continue;
3432                        else if (iter_mod > iter->module)
3433                                iter->tracepoint = NULL;
3434                        found = tracepoint_get_iter_range(&iter->tracepoint,
3435                                iter_mod->tracepoints_ptrs,
3436                                iter_mod->tracepoints_ptrs
3437                                        + iter_mod->num_tracepoints);
3438                        if (found) {
3439                                iter->module = iter_mod;
3440                                break;
3441                        }
3442                }
3443        }
3444        mutex_unlock(&module_mutex);
3445        return found;
3446}
3447#endif
3448