libexec/rtld-elf/malloc.c

   1 /*-
   2  * Copyright (c) 1983 Regents of the University of California.
   3  * All rights reserved.
   4  *
   5  * Redistribution and use in source and binary forms, with or without
   6  * modification, are permitted provided that the following conditions
   7  * are met:
   8  * 1. Redistributions of source code must retain the above copyright
   9  *    notice, this list of conditions and the following disclaimer.
  10  * 2. Redistributions in binary form must reproduce the above copyright
  11  *    notice, this list of conditions and the following disclaimer in the
  12  *    documentation and/or other materials provided with the distribution.
  13  * 3. Neither the name of the University nor the names of its contributors
  14  *    may be used to endorse or promote products derived from this software
  15  *    without specific prior written permission.
  16  *
  17  * THIS SOFTWARE IS PROVIDED BY THE REGENTS AND CONTRIBUTORS ``AS IS'' AND
  18  * ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
  19  * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
  20  * ARE DISCLAIMED.  IN NO EVENT SHALL THE REGENTS OR CONTRIBUTORS BE LIABLE
  21  * FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL
  22  * DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS
  23  * OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION)
  24  * HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
  25  * LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY
  26  * OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF
  27  * SUCH DAMAGE.
  28  */
  29
  30 /*
  31  * malloc.c (Caltech) 2/21/82
  32  * Chris Kingsley, kingsley@cit-20.
  33  *
  34  * This is a very fast storage allocator.  It allocates blocks of a small
  35  * number of different sizes, and keeps free lists of each size.  Blocks that
  36  * don't exactly fit are passed up to the next larger size.  In this
  37  * implementation, the available sizes are 2^n-4 (or 2^n-10) bytes long.
  38  * This is designed for use in a virtual memory environment.
  39  */
  40
  41 #include <sys/types.h>
  42 #include <paths.h>
  43 #include <stdarg.h>
  44 #include <stdio.h>
  45 #include <stdlib.h>
  46 #include <string.h>
  47 #include <unistd.h>
  48 #include <sys/param.h>
  49 #include <sys/mman.h>
  50 #include "rtld_printf.h"
  51
  52 static void morecore();
  53 static int findbucket();
  54
  55 /*
  56  * Pre-allocate mmap'ed pages
  57  */
  58 #define NPOOLPAGES      (32*1024/pagesz)
  59 static caddr_t          pagepool_start, pagepool_end;
  60 static int              morepages();
  61
  62 /*
  63  * The overhead on a block is at least 4 bytes.  When free, this space
  64  * contains a pointer to the next free block, and the bottom two bits must
  65  * be zero.  When in use, the first byte is set to MAGIC, and the second
  66  * byte is the size index.  The remaining bytes are for alignment.
  67  * If range checking is enabled then a second word holds the size of the
  68  * requested block, less 1, rounded up to a multiple of sizeof(RMAGIC).
  69  * The order of elements is critical: ov_magic must overlay the low order
  70  * bits of ov_next, and ov_magic can not be a valid ov_next bit pattern.
  71  */
  72 union   overhead {
  73         union   overhead *ov_next;      /* when free */
  74         struct {
  75                 u_char  ovu_magic;      /* magic number */
  76                 u_char  ovu_index;      /* bucket # */
  77 #ifdef RCHECK
  78                 u_short ovu_rmagic;     /* range magic number */
  79                 u_int   ovu_size;       /* actual block size */
  80 #endif
  81         } ovu;
  82 #define ov_magic        ovu.ovu_magic
  83 #define ov_index        ovu.ovu_index
  84 #define ov_rmagic       ovu.ovu_rmagic
  85 #define ov_size         ovu.ovu_size
  86 };
  87
  88 #define MAGIC           0xef            /* magic # on accounting info */
  89 #define RMAGIC          0x5555          /* magic # on range info */
  90
  91 #ifdef RCHECK
  92 #define RSLOP           sizeof (u_short)
  93 #else
  94 #define RSLOP           0
  95 #endif
  96
  97 /*
  98  * nextf[i] is the pointer to the next free block of size 2^(i+3).  The
  99  * smallest allocatable block is 8 bytes.  The overhead information
 100  * precedes the data area returned to the user.
 101  */
 102 #define NBUCKETS 30
 103 static  union overhead *nextf[NBUCKETS];
 104
 105 static  int pagesz;                     /* page size */
 106 static  int pagebucket;                 /* page size bucket */
 107
 108 #ifdef MSTATS
 109 /*
 110  * nmalloc[i] is the difference between the number of mallocs and frees
 111  * for a given block size.
 112  */
 113 static  u_int nmalloc[NBUCKETS];
 114 #include <stdio.h>
 115 #endif
 116
 117 #if defined(MALLOC_DEBUG) || defined(RCHECK)
 118 #define ASSERT(p)   if (!(p)) botch("p")
 119 #include <stdio.h>
 120 static void
 121 botch(char *s)
 122 {
 123         fprintf(stderr, "\r\nassertion botched: %s\r\n", s);
 124         (void) fflush(stderr);          /* just in case user buffered it */
 125         abort();
 126 }
 127 #else
 128 #define ASSERT(p)
 129 #endif
 130
 131 /* Debugging stuff */
 132 #define TRACE() rtld_printf("TRACE %s:%d\n", __FILE__, __LINE__)
 133
 134 void *
 135 malloc(size_t nbytes)
 136 {
 137         register union overhead *op;
 138         register int bucket;
 139         register long n;
 140         register unsigned amt;
 141
 142         /*
 143          * First time malloc is called, setup page size and
 144          * align break pointer so all data will be page aligned.
 145          */
 146         if (pagesz == 0) {
 147                 pagesz = n = getpagesize();
 148                 if (morepages(NPOOLPAGES) == 0)
 149                         return NULL;
 150                 op = (union overhead *)(pagepool_start);
 151                 n = n - sizeof (*op) - ((long)op & (n - 1));
 152                 if (n < 0)
 153                         n += pagesz;
 154                 if (n) {
 155                         pagepool_start += n;
 156                 }
 157                 bucket = 0;
 158                 amt = 8;
 159                 while ((unsigned)pagesz > amt) {
 160                         amt <<= 1;
 161                         bucket++;
 162                 }
 163                 pagebucket = bucket;
 164         }
 165         /*
 166          * Convert amount of memory requested into closest block size
 167          * stored in hash buckets which satisfies request.
 168          * Account for space used per block for accounting.
 169          */
 170         if (nbytes <= (unsigned long)(n = pagesz - sizeof (*op) - RSLOP)) {
 171 #ifndef RCHECK
 172                 amt = 8;        /* size of first bucket */
 173                 bucket = 0;
 174 #else
 175                 amt = 16;       /* size of first bucket */
 176                 bucket = 1;
 177 #endif
 178                 n = -(sizeof (*op) + RSLOP);
 179         } else {
 180                 amt = pagesz;
 181                 bucket = pagebucket;
 182         }
 183         while (nbytes > amt + n) {
 184                 amt <<= 1;
 185                 if (amt == 0)
 186                         return (NULL);
 187                 bucket++;
 188         }
 189         /*
 190          * If nothing in hash bucket right now,
 191          * request more memory from the system.
 192          */
 193         if ((op = nextf[bucket]) == NULL) {
 194                 morecore(bucket);
 195                 if ((op = nextf[bucket]) == NULL)
 196                         return (NULL);
 197         }
 198         /* remove from linked list */
 199         nextf[bucket] = op->ov_next;
 200         op->ov_magic = MAGIC;
 201         op->ov_index = bucket;
 202 #ifdef MSTATS
 203         nmalloc[bucket]++;
 204 #endif
 205 #ifdef RCHECK
 206         /*
 207          * Record allocated size of block and
 208          * bound space with magic numbers.
 209          */
 210         op->ov_size = (nbytes + RSLOP - 1) & ~(RSLOP - 1);
 211         op->ov_rmagic = RMAGIC;
 212         *(u_short *)((caddr_t)(op + 1) + op->ov_size) = RMAGIC;
 213 #endif
 214         return ((char *)(op + 1));
 215 }
 216
 217 /*
 218  * Used by rtld.c, if we don't override it here the calloc from
 219  * libc may try to pull in the malloc/realloc/free from libc too.
 220  */
 221 void *
 222 calloc(size_t num, size_t size)
 223 {
 224         void *p;
 225
 226         size *= num;
 227         if ((p = malloc(size)) != NULL)
 228                 bzero(p, size);
 229         return(p);
 230 }
 231
 232 /*
 233  * Allocate more memory to the indicated bucket.
 234  */
 235 static void
 236 morecore(int bucket)
 237 {
 238         register union overhead *op;
 239         register int sz;                /* size of desired block */
 240         int amt;                        /* amount to allocate */
 241         int nblks;                      /* how many blocks we get */
 242
 243         /*
 244          * sbrk_size <= 0 only for big, FLUFFY, requests (about
 245          * 2^30 bytes on a VAX, I think) or for a negative arg.
 246          */
 247         sz = 1 << (bucket + 3);
 248 #ifdef MALLOC_DEBUG
 249         ASSERT(sz > 0);
 250 #else
 251         if (sz <= 0)
 252                 return;
 253 #endif
 254         if (sz < pagesz) {
 255                 amt = pagesz;
 256                 nblks = amt / sz;
 257         } else {
 258                 amt = sz + pagesz;
 259                 nblks = 1;
 260         }
 261         if (amt > pagepool_end - pagepool_start)
 262                 if (morepages(amt/pagesz + NPOOLPAGES) == 0)
 263                         return;
 264         op = (union overhead *)pagepool_start;
 265         pagepool_start += amt;
 266
 267         /*
 268          * Add new memory allocated to that on
 269          * free list for this hash bucket.
 270          */
 271         nextf[bucket] = op;
 272         while (--nblks > 0) {
 273                 op->ov_next = (union overhead *)((caddr_t)op + sz);
 274                 op = (union overhead *)((caddr_t)op + sz);
 275         }
 276 }
 277
 278 void
 279 free(void *cp)
 280 {
 281         register int size;
 282         register union overhead *op;
 283
 284         if (cp == NULL)
 285                 return;
 286         op = (union overhead *)((caddr_t)cp - sizeof (union overhead));
 287 #ifdef MALLOC_DEBUG
 288         ASSERT(op->ov_magic == MAGIC);          /* make sure it was in use */
 289 #else
 290         if (op->ov_magic != MAGIC)
 291                 return;                         /* sanity */
 292 #endif
 293 #ifdef RCHECK
 294         ASSERT(op->ov_rmagic == RMAGIC);
 295         ASSERT(*(u_short *)((caddr_t)(op + 1) + op->ov_size) == RMAGIC);
 296 #endif
 297         size = op->ov_index;
 298         ASSERT(size < NBUCKETS);
 299         op->ov_next = nextf[size];      /* also clobbers ov_magic */
 300         nextf[size] = op;
 301 #ifdef MSTATS
 302         nmalloc[size]--;
 303 #endif
 304 }
 305
 306 /*
 307  * When a program attempts "storage compaction" as mentioned in the
 308  * old malloc man page, it realloc's an already freed block.  Usually
 309  * this is the last block it freed; occasionally it might be farther
 310  * back.  We have to search all the free lists for the block in order
 311  * to determine its bucket: 1st we make one pass thru the lists
 312  * checking only the first block in each; if that fails we search
 313  * ``realloc_srchlen'' blocks in each list for a match (the variable
 314  * is extern so the caller can modify it).  If that fails we just copy
 315  * however many bytes was given to realloc() and hope it's not huge.
 316  */
 317 int realloc_srchlen = 4;        /* 4 should be plenty, -1 =>'s whole list */
 318
 319 void *
 320 realloc(void *cp, size_t nbytes)
 321 {
 322         register u_int onb;
 323         register int i;
 324         union overhead *op;
 325         char *res;
 326         int was_alloced = 0;
 327
 328         if (cp == NULL)
 329                 return (malloc(nbytes));
 330         op = (union overhead *)((caddr_t)cp - sizeof (union overhead));
 331         if (op->ov_magic == MAGIC) {
 332                 was_alloced++;
 333                 i = op->ov_index;
 334         } else {
 335                 /*
 336                  * Already free, doing "compaction".
 337                  *
 338                  * Search for the old block of memory on the
 339                  * free list.  First, check the most common
 340                  * case (last element free'd), then (this failing)
 341                  * the last ``realloc_srchlen'' items free'd.
 342                  * If all lookups fail, then assume the size of
 343                  * the memory block being realloc'd is the
 344                  * largest possible (so that all "nbytes" of new
 345                  * memory are copied into).  Note that this could cause
 346                  * a memory fault if the old area was tiny, and the moon
 347                  * is gibbous.  However, that is very unlikely.
 348                  */
 349                 if ((i = findbucket(op, 1)) < 0 &&
 350                     (i = findbucket(op, realloc_srchlen)) < 0)
 351                         i = NBUCKETS;
 352         }
 353         onb = 1 << (i + 3);
 354         if (onb < (u_int)pagesz)
 355                 onb -= sizeof (*op) + RSLOP;
 356         else
 357                 onb += pagesz - sizeof (*op) - RSLOP;
 358         /* avoid the copy if same size block */
 359         if (was_alloced) {
 360                 if (i) {
 361                         i = 1 << (i + 2);
 362                         if (i < pagesz)
 363                                 i -= sizeof (*op) + RSLOP;
 364                         else
 365                                 i += pagesz - sizeof (*op) - RSLOP;
 366                 }
 367                 if (nbytes <= onb && nbytes > (size_t)i) {
 368 #ifdef RCHECK
 369                         op->ov_size = (nbytes + RSLOP - 1) & ~(RSLOP - 1);
 370                         *(u_short *)((caddr_t)(op + 1) + op->ov_size) = RMAGIC;
 371 #endif
 372                         return(cp);
 373                 } else
 374                         free(cp);
 375         }
 376         if ((res = malloc(nbytes)) == NULL)
 377                 return (NULL);
 378         if (cp != res)          /* common optimization if "compacting" */
 379                 bcopy(cp, res, (nbytes < onb) ? nbytes : onb);
 380         return (res);
 381 }
 382
 383 /*
 384  * Search ``srchlen'' elements of each free list for a block whose
 385  * header starts at ``freep''.  If srchlen is -1 search the whole list.
 386  * Return bucket number, or -1 if not found.
 387  */
 388 static int
 389 findbucket(union overhead *freep, int srchlen)
 390 {
 391         register union overhead *p;
 392         register int i, j;
 393
 394         for (i = 0; i < NBUCKETS; i++) {
 395                 j = 0;
 396                 for (p = nextf[i]; p && j != srchlen; p = p->ov_next) {
 397                         if (p == freep)
 398                                 return (i);
 399                         j++;
 400                 }
 401         }
 402         return (-1);
 403 }
 404
 405 #ifdef MSTATS
 406 /*
 407  * mstats - print out statistics about malloc
 408  *
 409  * Prints two lines of numbers, one showing the length of the free list
 410  * for each size category, the second showing the number of mallocs -
 411  * frees for each size category.
 412  */
 413 void
 414 mstats(char *s)
 415 {
 416         register int i, j;
 417         register union overhead *p;
 418         int totfree = 0,
 419         totused = 0;
 420
 421         fprintf(stderr, "Memory allocation statistics %s\nfree:\t", s);
 422         for (i = 0; i < NBUCKETS; i++) {
 423                 for (j = 0, p = nextf[i]; p; p = p->ov_next, j++)
 424                         ;
 425                 fprintf(stderr, " %d", j);
 426                 totfree += j * (1 << (i + 3));
 427         }
 428         fprintf(stderr, "\nused:\t");
 429         for (i = 0; i < NBUCKETS; i++) {
 430                 fprintf(stderr, " %d", nmalloc[i]);
 431                 totused += nmalloc[i] * (1 << (i + 3));
 432         }
 433         fprintf(stderr, "\n\tTotal in use: %d, total free: %d\n",
 434             totused, totfree);
 435 }
 436 #endif
 437
 438
 439 static int
 440 morepages(int n)
 441 {
 442         int     fd = -1;
 443         int     offset;
 444
 445         if (pagepool_end - pagepool_start > pagesz) {
 446                 caddr_t addr = (caddr_t)
 447                         (((long)pagepool_start + pagesz - 1) & ~(pagesz - 1));
 448                 if (munmap(addr, pagepool_end - addr) != 0)
 449                         rtld_fdprintf(STDERR_FILENO, "morepages: munmap %p",
 450                             addr);
 451         }
 452
 453         offset = (long)pagepool_start - ((long)pagepool_start & ~(pagesz - 1));
 454
 455         if ((pagepool_start = mmap(0, n * pagesz,
 456                         PROT_READ|PROT_WRITE,
 457                         MAP_ANON|MAP_COPY, fd, 0)) == (caddr_t)-1) {
 458                 rtld_printf("Cannot map anonymous memory\n");
 459                 return 0;
 460         }
 461         pagepool_end = pagepool_start + n * pagesz;
 462         pagepool_start += offset;
 463
 464         return n;
 465 }