binutils-gdb/gdb/objfiles.h
Fred Fish bf349b77fa Add OBJF_SYMS flag and use it to decide whether or not we have already
tried to read symbols from a file, for mapped objfiles.  This fixes a
memory leak (actually a constant memory growth) due to reading symbol
files with no debug info and thus generating no psymtabs or symtabs.
Most typically happened with shared libraries.
1992-03-28 21:06:23 +00:00

323 lines
13 KiB
C

/* Definitions for symbol file management in GDB.
Copyright (C) 1992 Free Software Foundation, Inc.
This file is part of GDB.
This program is free software; you can redistribute it and/or modify
it under the terms of the GNU General Public License as published by
the Free Software Foundation; either version 2 of the License, or
(at your option) any later version.
This program is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
GNU General Public License for more details.
You should have received a copy of the GNU General Public License
along with this program; if not, write to the Free Software
Foundation, Inc., 675 Mass Ave, Cambridge, MA 02139, USA. */
#if !defined (OBJFILES_H)
#define OBJFILES_H
/* This structure maintains information on a per-objfile basis about the
"entry point" of the objfile, and the scope within which the entry point
exists. It is possible that gdb will see more than one objfile that is
executable, each with it's own entry point.
For example, for dynamically linked executables in SVR4, the dynamic linker
code is contained within the shared C library, which is actually executable
and is run by the kernel first when an exec is done of a user executable
that is dynamically linked. The dynamic linker within the shared C library
then maps in the various program segments in the user executable and jumps
to the user executable's recorded entry point, as if the call had been made
directly by the kernel.
The traditional gdb method of using this info is to use the recorded entry
point to set the variables entry_file_lowpc and entry_file_highpc from
the debugging information, where these values are the starting address
(inclusive) and ending address (exclusive) of the instruction space in the
executable which correspond to the "startup file", I.E. crt0.o in most
cases. This file is assumed to be a startup file and frames with pc's
inside it are treated as nonexistent. Setting these variables is necessary
so that backtraces do not fly off the bottom of the stack (or top, depending
upon your stack orientation).
Gdb also supports an alternate method to avoid running off the top/bottom
of the stack.
There are two frames that are "special", the frame for the function
containing the process entry point, since it has no predecessor frame,
and the frame for the function containing the user code entry point
(the main() function), since all the predecessor frames are for the
process startup code. Since we have no guarantee that the linked
in startup modules have any debugging information that gdb can use,
we need to avoid following frame pointers back into frames that might
have been built in the startup code, as we might get hopelessly
confused. However, we almost always have debugging information
available for main().
These variables are used to save the range of PC values which are valid
within the main() function and within the function containing the process
entry point. If we always consider the frame for main() as the outermost
frame when debugging user code, and the frame for the process entry
point function as the outermost frame when debugging startup code, then
all we have to do is have FRAME_CHAIN_VALID return false whenever a
frame's current PC is within the range specified by these variables.
In essence, we set "ceilings" in the frame chain beyond which we will
not proceed when following the frame chain back up the stack.
A nice side effect is that we can still debug startup code without
running off the end of the frame chain, assuming that we have usable
debugging information in the startup modules, and if we choose to not
use the block at main, or can't find it for some reason, everything
still works as before. And if we have no startup code debugging
information but we do have usable information for main(), backtraces
from user code don't go wandering off into the startup code.
To use this method, define your FRAME_CHAIN_VALID macro like:
#define FRAME_CHAIN_VALID(chain, thisframe) \
(chain != 0 \
&& !(inside_main_func ((thisframe)->pc)) \
&& !(inside_entry_func ((thisframe)->pc)))
and add initializations of the four scope controlling variables inside
the object file / debugging information processing modules. */
struct entry_info
{
/* The value we should use for this objects entry point.
The illegal/unknown value needs to be something other than 0, ~0
for instance, which is much less likely than 0. */
CORE_ADDR entry_point;
/* Start (inclusive) and end (exclusive) of function containing the
entry point. */
CORE_ADDR entry_func_lowpc;
CORE_ADDR entry_func_highpc;
/* Start (inclusive) and end (exclusive) of object file containing the
entry point. */
CORE_ADDR entry_file_lowpc;
CORE_ADDR entry_file_highpc;
/* Start (inclusive) and end (exclusive) of the user code main() function. */
CORE_ADDR main_func_lowpc;
CORE_ADDR main_func_highpc;
};
/* Master structure for keeping track of each input file from which
gdb reads symbols. One of these is allocated for each such file we
access, e.g. the exec_file, symbol_file, and any shared library object
files. */
struct objfile
{
/* All struct objfile's are chained together by their next pointers.
The global variable "object_files" points to the first link in this
chain. */
struct objfile *next;
/* The object file's name. Malloc'd; free it if you free this struct. */
char *name;
/* Some flag bits for this objfile. */
unsigned short flags;
/* Each objfile points to a linked list of symtabs derived from this file,
one symtab structure for each compilation unit (source file). Each link
in the symtab list contains a backpointer to this objfile. */
struct symtab *symtabs;
/* Each objfile points to a linked list of partial symtabs derived from
this file, one partial symtab structure for each compilation unit
(source file). */
struct partial_symtab *psymtabs;
/* List of freed partial symtabs, available for re-use */
struct partial_symtab *free_psymtabs;
/* The object file's BFD. Can be null, in which case bfd_open (name) and
put the result here. */
bfd *obfd;
/* The modification timestamp of the object file, as of the last time
we read its symbols. */
long mtime;
/* Obstacks to hold objects that should be freed when we load a new symbol
table from this object file. */
struct obstack psymbol_obstack; /* Partial symbols */
struct obstack symbol_obstack; /* Full symbols */
struct obstack type_obstack; /* Types */
/* Vectors of all partial symbols read in from file. The actual data
is stored in the psymbol_obstack. */
struct psymbol_allocation_list global_psymbols;
struct psymbol_allocation_list static_psymbols;
/* Each file contains a pointer to an array of minimal symbols for all
global symbols that are defined within the file. The array is terminated
by a "null symbol", one that has a NULL pointer for the name and a zero
value for the address. This makes it easy to walk through the array
when passed a pointer to somewhere in the middle of it. There is also
a count of the number of symbols, which does include the terminating
null symbol. The array itself, as well as all the data that it points
to, should be allocated on the symbol_obstack for this file. */
struct minimal_symbol *msymbols;
int minimal_symbol_count;
/* For object file formats which don't specify fundamental types, gdb
can create such types. For now, it maintains a vector of pointers
to these internally created fundamental types on a per objfile basis,
however it really should ultimately keep them on a per-compilation-unit
basis, to account for linkage-units that consist of a number of
compilation units that may have different fundamental types, such as
linking C modules with ADA modules, or linking C modules that are
compiled with 32-bit ints with C modules that are compiled with 64-bit
ints (not inherently evil with a smarter linker). */
struct type **fundamental_types;
/* The mmalloc() malloc-descriptor for this objfile if we are using
the memory mapped malloc() package to manage storage for this objfile's
data. NULL if we are not. */
PTR md;
/* Structure which keeps track of functions that manipulate objfile's
of the same type as this objfile. I.E. the function to read partial
symbols for example. Note that this structure is in statically
allocated memory, and is shared by all objfiles that use the
object module reader of this type. */
struct sym_fns *sf;
/* The per-objfile information about the entry point, the scope (file/func)
containing the entry point, and the scope of the user's main() func. */
struct entry_info ei;
/* Hook for information which is shared by sym_init and sym_read for
this objfile. It is typically a pointer to malloc'd memory. */
PTR sym_private;
};
/* Defines for the objfile flag word. */
/* Gdb can arrange to allocate storage for all objects related to a
particular objfile in a designated section of it's address space,
managed at a low level by mmap() and using a special version of
malloc that handles malloc/free/realloc on top of the mmap() interface.
This allows the "internal gdb state" for a particular objfile to be
dumped to a gdb state file and subsequently reloaded at a later time. */
#define OBJF_MAPPED (1 << 0) /* Objfile data is mmap'd */
/* When using mapped/remapped predigested gdb symbol information, we need
a flag that indicates that we have previously done an initial symbol
table read from this particular objfile. We can't just look for the
absence of any of the three symbol tables (msymbols, psymtab, symtab)
because if the file has no symbols for example, none of these will
exist. */
#define OBJF_SYMS (1 << 1) /* Have tried to read symbols */
/* The object file that the main symbol table was loaded from (e.g. the
argument to the "symbol-file" or "file" command). */
extern struct objfile *symfile_objfile;
/* When we need to allocate a new type, we need to know which type_obstack
to allocate the type on, since there is one for each objfile. The places
where types are allocated are deeply buried in function call hierarchies
which know nothing about objfiles, so rather than trying to pass a
particular objfile down to them, we just do an end run around them and
set current_objfile to be whatever objfile we expect to be using at the
time types are being allocated. For instance, when we start reading
symbols for a particular objfile, we set current_objfile to point to that
objfile, and when we are done, we set it back to NULL, to ensure that we
never put a type someplace other than where we are expecting to put it.
FIXME: Maybe we should review the entire type handling system and
see if there is a better way to avoid this problem. */
extern struct objfile *current_objfile;
/* All known objfiles are kept in a linked list. This points to the
root of this list. */
extern struct objfile *object_files;
/* Declarations for functions defined in objfiles.c */
extern struct objfile *
allocate_objfile PARAMS ((bfd *, int));
extern void
free_objfile PARAMS ((struct objfile *));
extern void
free_all_objfiles PARAMS ((void));
extern int
have_partial_symbols PARAMS ((void));
extern int
have_full_symbols PARAMS ((void));
/* Functions for dealing with the minimal symbol table, really a misc
address<->symbol mapping for things we don't have debug symbols for. */
extern int
have_minimal_symbols PARAMS ((void));
extern PTR
iterate_over_objfiles PARAMS ((PTR (*func) (struct objfile *,
PTR arg1, PTR arg2, PTR arg3),
PTR arg1, PTR arg2, PTR arg3));
extern PTR
iterate_over_symtabs PARAMS ((PTR (*func) (struct objfile *, struct symtab *,
PTR arg1, PTR arg2, PTR arg3),
PTR arg1, PTR arg2, PTR arg3));
extern PTR
iterate_over_psymtabs PARAMS ((PTR (*func) (struct objfile *,
struct partial_symtab *,
PTR arg1, PTR arg2, PTR arg3),
PTR arg1, PTR arg2, PTR arg3));
/* Traverse all object files. ALL_OBJFILES_SAFE works even if you delete
the objfile during the traversal. */
#define ALL_OBJFILES(obj) \
for ((obj)=object_files; (obj)!=NULL; (obj)=(obj)->next)
#define ALL_OBJFILES_SAFE(obj,nxt) \
for ((obj)=object_files; (obj)!=NULL?((nxt)=(obj)->next,1):0; (obj)=(nxt))
#endif /* !defined (OBJFILES_H) */