gdb/minsyms.c

   1 /* GDB routines for manipulating the minimal symbol tables.
   2    Copyright 1992 Free Software Foundation, Inc.
   3    Contributed by Cygnus Support, using pieces from other GDB modules.
   4
   5 This file is part of GDB.
   6
   7 This program is free software; you can redistribute it and/or modify
   8 it under the terms of the GNU General Public License as published by
   9 the Free Software Foundation; either version 2 of the License, or
  10 (at your option) any later version.
  11
  12 This program is distributed in the hope that it will be useful,
  13 but WITHOUT ANY WARRANTY; without even the implied warranty of
  14 MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
  15 GNU General Public License for more details.
  16
  17 You should have received a copy of the GNU General Public License
  18 along with this program; if not, write to the Free Software
  19 Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA.  */
  20
  21
  22 /* This file contains support routines for creating, manipulating, and
  23    destroying minimal symbol tables.
  24
  25    Minimal symbol tables are used to hold some very basic information about
  26    all defined global symbols (text, data, bss, abs, etc).  The only two
  27    required pieces of information are the symbol's name and the address
  28    associated with that symbol.
  29
  30    In many cases, even if a file was compiled with no special options for
  31    debugging at all, as long as was not stripped it will contain sufficient
  32    information to build useful minimal symbol tables using this structure.
  33
  34    Even when a file contains enough debugging information to build a full
  35    symbol table, these minimal symbols are still useful for quickly mapping
  36    between names and addresses, and vice versa.  They are also sometimes used
  37    to figure out what full symbol table entries need to be read in. */
  38
  39
  40 #include "defs.h"
  41 #include "symtab.h"
  42 #include "bfd.h"
  43 #include "symfile.h"
  44 #include "objfiles.h"
  45 #include "demangle.h"
  46
  47 /* Accumulate the minimal symbols for each objfile in bunches of BUNCH_SIZE.
  48    At the end, copy them all into one newly allocated location on an objfile's
  49    symbol obstack.  */
  50
  51 #define BUNCH_SIZE 127
  52
  53 struct msym_bunch
  54 {
  55   struct msym_bunch *next;
  56   struct minimal_symbol contents[BUNCH_SIZE];
  57 };
  58
  59 /* Bunch currently being filled up.
  60    The next field points to chain of filled bunches.  */
  61
  62 static struct msym_bunch *msym_bunch;
  63
  64 /* Number of slots filled in current bunch.  */
  65
  66 static int msym_bunch_index;
  67
  68 /* Total number of minimal symbols recorded so far for the objfile.  */
  69
  70 static int msym_count;
  71
  72 /* Prototypes for local functions. */
  73
  74 static int
  75 compare_minimal_symbols PARAMS ((const void *, const void *));
  76
  77 static int
  78 compact_minimal_symbols PARAMS ((struct minimal_symbol *, int));
  79
  80 /* Look through all the current minimal symbol tables and find the first
  81    minimal symbol that matches NAME.  If OBJF is non-NULL, it specifies a
  82    particular objfile and the search is limited to that objfile.  Returns
  83    a pointer to the minimal symbol that matches, or NULL if no match is found.
  84
  85    Note:  One instance where there may be duplicate minimal symbols with
  86    the same name is when the symbol tables for a shared library and the
  87    symbol tables for an executable contain global symbols with the same
  88    names (the dynamic linker deals with the duplication). */
  89
  90 struct minimal_symbol *
  91 lookup_minimal_symbol (name, objf)
  92      register const char *name;
  93      struct objfile *objf;
  94 {
  95   struct objfile *objfile;
  96   struct minimal_symbol *msymbol;
  97   struct minimal_symbol *found_symbol = NULL;
  98 #ifdef IBM6000_TARGET
  99   struct minimal_symbol *trampoline_symbol = NULL;
 100 #endif
 101
 102   for (objfile = object_files;
 103        objfile != NULL && found_symbol == NULL;
 104        objfile = objfile -> next)
 105     {
 106       if (objf == NULL || objf == objfile)
 107         {
 108           for (msymbol = objfile -> msymbols;
 109                msymbol != NULL && SYMBOL_NAME (msymbol) != NULL &&
 110                found_symbol == NULL;
 111                msymbol++)
 112             {
 113               if (SYMBOL_MATCHES_NAME (msymbol, name))
 114                 {
 115 #ifdef IBM6000_TARGET
 116                   /* I *think* all platforms using shared libraries (and
 117                      trampoline code) will suffer this problem. Consider a
 118                      case where there are 5 shared libraries, each referencing
 119                      `foo' with a trampoline entry. When someone wants to put
 120                      a breakpoint on `foo' and the only info we have is minimal
 121                      symbol vector, we want to use the real `foo', rather than
 122                      one of those trampoline entries. MGO */
 123                   /* If a trampoline symbol is found, we prefer to keep looking
 124                      for the *real* symbol. If the actual symbol not found,
 125                      then we'll use the trampoline entry. Sorry for the machine
 126                      dependent code here, but I hope this will benefit other
 127                      platforms as well. For trampoline entries, we used
 128                      mst_unknown earlier. Perhaps we should define a
 129                      `mst_trampoline' type?? */
 130
 131                   if (MSYMBOL_TYPE (msymbol) != mst_unknown)
 132                     found_symbol = msymbol;
 133                   else if (MSYMBOL_TYPE (msymbol) == mst_unknown &&
 134                            !trampoline_symbol)
 135                     trampoline_symbol = msymbol;
 136
 137 #else
 138                   found_symbol = msymbol;
 139 #endif
 140                 }
 141             }
 142         }
 143     }
 144 #ifdef IBM6000_TARGET
 145   return found_symbol ? found_symbol : trampoline_symbol;
 146 #endif
 147
 148   return (found_symbol);
 149 }
 150
 151
 152 /* Search through the minimal symbol table for each objfile and find the
 153    symbol whose address is the largest address that is still less than or
 154    equal to PC.  Returns a pointer to the minimal symbol if such a symbol
 155    is found, or NULL if PC is not in a suitable range.  Note that we need
 156    to look through ALL the minimal symbol tables before deciding on the
 157    symbol that comes closest to the specified PC. */
 158
 159 struct minimal_symbol *
 160 lookup_minimal_symbol_by_pc (pc)
 161      register CORE_ADDR pc;
 162 {
 163   register int lo;
 164   register int hi;
 165   register int new;
 166   register struct objfile *objfile;
 167   register struct minimal_symbol *msymbol;
 168   register struct minimal_symbol *best_symbol = NULL;
 169
 170   for (objfile = object_files;
 171        objfile != NULL;
 172        objfile = objfile -> next)
 173     {
 174       /* If this objfile has a minimal symbol table, go search it using
 175          a binary search.  Note that a minimal symbol table always consists
 176          of at least two symbols, a "real" symbol and the terminating
 177          "null symbol".  If there are no real symbols, then there is no
 178          minimal symbol table at all. */
 179
 180       if ((msymbol = objfile -> msymbols) != NULL)
 181         {
 182           lo = 0;
 183           hi = objfile -> minimal_symbol_count - 1;
 184
 185           /* This code assumes that the minimal symbols are sorted by
 186              ascending address values.  If the pc value is greater than or
 187              equal to the first symbol's address, then some symbol in this
 188              minimal symbol table is a suitable candidate for being the
 189              "best" symbol.  This includes the last real symbol, for cases
 190              where the pc value is larger than any address in this vector.
 191
 192              By iterating until the address associated with the current
 193              hi index (the endpoint of the test interval) is less than
 194              or equal to the desired pc value, we accomplish two things:
 195              (1) the case where the pc value is larger than any minimal
 196              symbol address is trivially solved, (2) the address associated
 197              with the hi index is always the one we want when the interation
 198              terminates.  In essence, we are iterating the test interval
 199              down until the pc value is pushed out of it from the high end.
 200
 201              Warning: this code is trickier than it would appear at first. */
 202
 203           /* Should also requires that pc is <= end of objfile.  FIXME! */
 204           if (pc >= SYMBOL_VALUE_ADDRESS (&msymbol[lo]))
 205             {
 206               while (SYMBOL_VALUE_ADDRESS (&msymbol[hi]) > pc)
 207                 {
 208                   /* pc is still strictly less than highest address */
 209                   /* Note "new" will always be >= lo */
 210                   new = (lo + hi) / 2;
 211                   if ((SYMBOL_VALUE_ADDRESS (&msymbol[new]) >= pc) ||
 212                       (lo == new))
 213                     {
 214                       hi = new;
 215                     }
 216                   else
 217                     {
 218                       lo = new;
 219                     }
 220                 }
 221               /* The minimal symbol indexed by hi now is the best one in this
 222                  objfile's minimal symbol table.  See if it is the best one
 223                  overall. */
 224
 225               if ((best_symbol == NULL) ||
 226                   (SYMBOL_VALUE_ADDRESS (best_symbol) <
 227                    SYMBOL_VALUE_ADDRESS (&msymbol[hi])))
 228                 {
 229                   best_symbol = &msymbol[hi];
 230                 }
 231             }
 232         }
 233     }
 234   return (best_symbol);
 235 }
 236
 237 /* Prepare to start collecting minimal symbols.  Note that presetting
 238    msym_bunch_index to BUNCH_SIZE causes the first call to save a minimal
 239    symbol to allocate the memory for the first bunch. */
 240
 241 void
 242 init_minimal_symbol_collection ()
 243 {
 244   msym_count = 0;
 245   msym_bunch = NULL;
 246   msym_bunch_index = BUNCH_SIZE;
 247 }
 248
 249 void
 250 prim_record_minimal_symbol (name, address, ms_type)
 251      const char *name;
 252      CORE_ADDR address;
 253      enum minimal_symbol_type ms_type;
 254 {
 255   register struct msym_bunch *new;
 256   register struct minimal_symbol *msymbol;
 257
 258   if (msym_bunch_index == BUNCH_SIZE)
 259     {
 260       new = (struct msym_bunch *) xmalloc (sizeof (struct msym_bunch));
 261       msym_bunch_index = 0;
 262       new -> next = msym_bunch;
 263       msym_bunch = new;
 264     }
 265   msymbol = &msym_bunch -> contents[msym_bunch_index];
 266   SYMBOL_NAME (msymbol) = (char *) name;
 267   SYMBOL_INIT_LANGUAGE_SPECIFIC (msymbol, language_unknown);
 268   SYMBOL_VALUE_ADDRESS (msymbol) = address;
 269   MSYMBOL_TYPE (msymbol) = ms_type;
 270   /* FIXME:  This info, if it remains, needs its own field.  */
 271   MSYMBOL_INFO (msymbol) = NULL; /* FIXME! */
 272   msym_bunch_index++;
 273   msym_count++;
 274 }
 275
 276 /* FIXME:  Why don't we just combine this function with the one above
 277    and pass it a NULL info pointer value if info is not needed? */
 278
 279 void
 280 prim_record_minimal_symbol_and_info (name, address, ms_type, info)
 281      const char *name;
 282      CORE_ADDR address;
 283      enum minimal_symbol_type ms_type;
 284      char *info;
 285 {
 286   register struct msym_bunch *new;
 287   register struct minimal_symbol *msymbol;
 288
 289   if (msym_bunch_index == BUNCH_SIZE)
 290     {
 291       new = (struct msym_bunch *) xmalloc (sizeof (struct msym_bunch));
 292       msym_bunch_index = 0;
 293       new -> next = msym_bunch;
 294       msym_bunch = new;
 295     }
 296   msymbol = &msym_bunch -> contents[msym_bunch_index];
 297   SYMBOL_NAME (msymbol) = (char *) name;
 298   SYMBOL_INIT_LANGUAGE_SPECIFIC (msymbol, language_unknown);
 299   SYMBOL_VALUE_ADDRESS (msymbol) = address;
 300   MSYMBOL_TYPE (msymbol) = ms_type;
 301   /* FIXME:  This info, if it remains, needs its own field.  */
 302   MSYMBOL_INFO (msymbol) = info; /* FIXME! */
 303   msym_bunch_index++;
 304   msym_count++;
 305 }
 306
 307 /* Compare two minimal symbols by address and return a signed result based
 308    on unsigned comparisons, so that we sort into unsigned numeric order.  */
 309
 310 static int
 311 compare_minimal_symbols (fn1p, fn2p)
 312      const PTR fn1p;
 313      const PTR fn2p;
 314 {
 315   register const struct minimal_symbol *fn1;
 316   register const struct minimal_symbol *fn2;
 317
 318   fn1 = (const struct minimal_symbol *) fn1p;
 319   fn2 = (const struct minimal_symbol *) fn2p;
 320
 321   if (SYMBOL_VALUE_ADDRESS (fn1) < SYMBOL_VALUE_ADDRESS (fn2))
 322     {
 323       return (-1);
 324     }
 325   else if (SYMBOL_VALUE_ADDRESS (fn1) > SYMBOL_VALUE_ADDRESS (fn2))
 326     {
 327       return (1);
 328     }
 329   else
 330     {
 331       return (0);
 332     }
 333 }
 334
 335 /* Discard the currently collected minimal symbols, if any.  If we wish
 336    to save them for later use, we must have already copied them somewhere
 337    else before calling this function.
 338
 339    FIXME:  We could allocate the minimal symbol bunches on their own
 340    obstack and then simply blow the obstack away when we are done with
 341    it.  Is it worth the extra trouble though? */
 342
 343 /* ARGSUSED */
 344 void
 345 discard_minimal_symbols (foo)
 346      int foo;
 347 {
 348   register struct msym_bunch *next;
 349
 350   while (msym_bunch != NULL)
 351     {
 352       next = msym_bunch -> next;
 353       free ((PTR)msym_bunch);
 354       msym_bunch = next;
 355     }
 356 }
 357
 358 /* Compact duplicate entries out of a minimal symbol table by walking
 359    through the table and compacting out entries with duplicate addresses
 360    and matching names.  Return the number of entries remaining.
 361
 362    On entry, the table resides between msymbol[0] and msymbol[mcount].
 363    On exit, it resides between msymbol[0] and msymbol[result_count].
 364
 365    When files contain multiple sources of symbol information, it is
 366    possible for the minimal symbol table to contain many duplicate entries.
 367    As an example, SVR4 systems use ELF formatted object files, which
 368    usually contain at least two different types of symbol tables (a
 369    standard ELF one and a smaller dynamic linking table), as well as
 370    DWARF debugging information for files compiled with -g.
 371
 372    Without compacting, the minimal symbol table for gdb itself contains
 373    over a 1000 duplicates, about a third of the total table size.  Aside
 374    from the potential trap of not noticing that two successive entries
 375    identify the same location, this duplication impacts the time required
 376    to linearly scan the table, which is done in a number of places.  So we
 377    just do one linear scan here and toss out the duplicates.
 378
 379    Note that we are not concerned here about recovering the space that
 380    is potentially freed up, because the strings themselves are allocated
 381    on the symbol_obstack, and will get automatically freed when the symbol
 382    table is freed.  The caller can free up the unused minimal symbols at
 383    the end of the compacted region if their allocation strategy allows it.
 384
 385    Also note we only go up to the next to last entry within the loop
 386    and then copy the last entry explicitly after the loop terminates.
 387
 388    Since the different sources of information for each symbol may
 389    have different levels of "completeness", we may have duplicates
 390    that have one entry with type "mst_unknown" and the other with a
 391    known type.  So if the one we are leaving alone has type mst_unknown,
 392    overwrite its type with the type from the one we are compacting out.  */
 393
 394 static int
 395 compact_minimal_symbols (msymbol, mcount)
 396      struct minimal_symbol *msymbol;
 397      int mcount;
 398 {
 399   struct minimal_symbol *copyfrom;
 400   struct minimal_symbol *copyto;
 401
 402   if (mcount > 0)
 403     {
 404       copyfrom = copyto = msymbol;
 405       while (copyfrom < msymbol + mcount - 1)
 406         {
 407           if (SYMBOL_VALUE_ADDRESS (copyfrom) ==
 408               SYMBOL_VALUE_ADDRESS ((copyfrom + 1)) &&
 409               (STREQ (SYMBOL_NAME (copyfrom), SYMBOL_NAME ((copyfrom + 1)))))
 410             {
 411               if (MSYMBOL_TYPE((copyfrom + 1)) == mst_unknown)
 412                 {
 413                   MSYMBOL_TYPE ((copyfrom + 1)) = MSYMBOL_TYPE (copyfrom);
 414                 }
 415               copyfrom++;
 416             }
 417           else
 418             {
 419               *copyto++ = *copyfrom++;
 420             }
 421         }
 422       *copyto++ = *copyfrom++;
 423       mcount = copyto - msymbol;
 424     }
 425   return (mcount);
 426 }
 427
 428 /* Add the minimal symbols in the existing bunches to the objfile's official
 429    minimal symbol table.  In most cases there is no minimal symbol table yet
 430    for this objfile, and the existing bunches are used to create one.  Once
 431    in a while (for shared libraries for example), we add symbols (e.g. common
 432    symbols) to an existing objfile.
 433
 434    Because of the way minimal symbols are collected, we generally have no way
 435    of knowing what source language applies to any particular minimal symbol.
 436    Specifically, we have no way of knowing if the minimal symbol comes from a
 437    C++ compilation unit or not.  So for the sake of supporting cached
 438    demangled C++ names, we have no choice but to try and demangle each new one
 439    that comes in.  If the demangling succeeds, then we assume it is a C++
 440    symbol and set the symbol's language and demangled name fields
 441    appropriately.  Note that in order to avoid unnecessary demanglings, and
 442    allocating obstack space that subsequently can't be freed for the demangled
 443    names, we mark all newly added symbols with language_auto.  After
 444    compaction of the minimal symbols, we go back and scan the entire minimal
 445    symbol table looking for these new symbols.  For each new symbol we attempt
 446    to demangle it, and if successful, record it as a language_cplus symbol
 447    and cache the demangled form on the symbol obstack.  Symbols which don't
 448    demangle are marked as language_unknown symbols, which inhibits future
 449    attempts to demangle them if we later add more minimal symbols. */
 450
 451 void
 452 install_minimal_symbols (objfile)
 453      struct objfile *objfile;
 454 {
 455   register int bindex;
 456   register int mcount;
 457   register struct msym_bunch *bunch;
 458   register struct minimal_symbol *msymbols;
 459   int alloc_count;
 460   register char leading_char;
 461   char *demangled_name;
 462
 463   if (msym_count > 0)
 464     {
 465       /* Allocate enough space in the obstack, into which we will gather the
 466          bunches of new and existing minimal symbols, sort them, and then
 467          compact out the duplicate entries.  Once we have a final table,
 468          we will give back the excess space.  */
 469
 470       alloc_count = msym_count + objfile->minimal_symbol_count + 1;
 471       obstack_blank (&objfile->symbol_obstack,
 472                      alloc_count * sizeof (struct minimal_symbol));
 473       msymbols = (struct minimal_symbol *)
 474                  obstack_base (&objfile->symbol_obstack);
 475
 476       /* Copy in the existing minimal symbols, if there are any.  */
 477
 478       if (objfile->minimal_symbol_count)
 479         memcpy ((char *)msymbols, (char *)objfile->msymbols,
 480                 objfile->minimal_symbol_count * sizeof (struct minimal_symbol));
 481
 482       /* Walk through the list of minimal symbol bunches, adding each symbol
 483          to the new contiguous array of symbols.  Note that we start with the
 484          current, possibly partially filled bunch (thus we use the current
 485          msym_bunch_index for the first bunch we copy over), and thereafter
 486          each bunch is full. */
 487
 488       mcount = objfile->minimal_symbol_count;
 489       leading_char = bfd_get_symbol_leading_char (objfile->obfd);
 490
 491       for (bunch = msym_bunch; bunch != NULL; bunch = bunch -> next)
 492         {
 493           for (bindex = 0; bindex < msym_bunch_index; bindex++, mcount++)
 494             {
 495               msymbols[mcount] = bunch -> contents[bindex];
 496               SYMBOL_LANGUAGE (&msymbols[mcount]) = language_auto;
 497               if (SYMBOL_NAME (&msymbols[mcount])[0] == leading_char)
 498                 {
 499                   SYMBOL_NAME(&msymbols[mcount])++;
 500                 }
 501             }
 502           msym_bunch_index = BUNCH_SIZE;
 503         }
 504
 505       /* Sort the minimal symbols by address.  */
 506
 507       qsort (msymbols, mcount, sizeof (struct minimal_symbol),
 508              compare_minimal_symbols);
 509
 510       /* Compact out any duplicates, and free up whatever space we are
 511          no longer using.  */
 512
 513       mcount = compact_minimal_symbols (msymbols, mcount);
 514
 515       obstack_blank (&objfile->symbol_obstack,
 516         (mcount + 1 - alloc_count) * sizeof (struct minimal_symbol));
 517       msymbols = (struct minimal_symbol *)
 518         obstack_finish (&objfile->symbol_obstack);
 519
 520       /* We also terminate the minimal symbol table with a "null symbol",
 521          which is *not* included in the size of the table.  This makes it
 522          easier to find the end of the table when we are handed a pointer
 523          to some symbol in the middle of it.  Zero out the fields in the
 524          "null symbol" allocated at the end of the array.  Note that the
 525          symbol count does *not* include this null symbol, which is why it
 526          is indexed by mcount and not mcount-1. */
 527
 528       SYMBOL_NAME (&msymbols[mcount]) = NULL;
 529       SYMBOL_VALUE_ADDRESS (&msymbols[mcount]) = 0;
 530       MSYMBOL_INFO (&msymbols[mcount]) = NULL;
 531       MSYMBOL_TYPE (&msymbols[mcount]) = mst_unknown;
 532       SYMBOL_INIT_LANGUAGE_SPECIFIC (&msymbols[mcount], language_unknown);
 533
 534       /* Attach the minimal symbol table to the specified objfile.
 535          The strings themselves are also located in the symbol_obstack
 536          of this objfile.  */
 537
 538       objfile -> minimal_symbol_count = mcount;
 539       objfile -> msymbols = msymbols;
 540
 541       /* Now walk through all the minimal symbols, selecting the newly added
 542          ones and attempting to cache their C++ demangled names. */
 543
 544       for ( ; mcount-- > 0 ; msymbols++)
 545         {
 546           SYMBOL_INIT_DEMANGLED_NAME (msymbols, &objfile->symbol_obstack);
 547         }
 548     }
 549 }
 550