/* Definitions for symbol file management in GDB.
Copyright (C) 1992-2019 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 3 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, see . */
#if !defined (OBJFILES_H)
#define OBJFILES_H
#include "hashtab.h"
#include "gdb_obstack.h" /* For obstack internals. */
#include "objfile-flags.h"
#include "symfile.h"
#include "progspace.h"
#include "registry.h"
#include "gdb_bfd.h"
#include
#include "common/next-iterator.h"
#include "common/safe-iterator.h"
struct bcache;
struct htab;
struct objfile_data;
struct partial_symbol;
/* 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 its 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 was to use the
recorded entry point to set the entry-file's lowpc and 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.
NOTE: cagney/2003-09-09: It turns out that this "traditional"
method doesn't work. Corinna writes: ``It turns out that the call
to test for "inside entry file" destroys a meaningful backtrace
under some conditions. E.g. the backtrace tests in the asm-source
testcase are broken for some targets. In this test the functions
are all implemented as part of one file and the testcase is not
necessarily linked with a start file (depending on the target).
What happens is, that the first frame is printed normaly and
following frames are treated as being inside the enttry file then.
This way, only the #0 frame is printed in the backtrace output.''
Ref "frame.c" "NOTE: vinschen/2003-04-01".
Gdb also supports an alternate method to avoid running off the 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
DEPRECATED_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. */
struct entry_info
{
/* The unrelocated value we should use for this objfile entry point. */
CORE_ADDR entry_point;
/* The index of the section in which the entry point appears. */
int the_bfd_section_index;
/* Set to 1 iff ENTRY_POINT contains a valid value. */
unsigned entry_point_p : 1;
/* Set to 1 iff this object was initialized. */
unsigned initialized : 1;
};
/* Sections in an objfile. The section offsets are stored in the
OBJFILE. */
struct obj_section
{
/* BFD section pointer */
struct bfd_section *the_bfd_section;
/* Objfile this section is part of. */
struct objfile *objfile;
/* True if this "overlay section" is mapped into an "overlay region". */
int ovly_mapped;
};
/* Relocation offset applied to S. */
#define obj_section_offset(s) \
(((s)->objfile->section_offsets)->offsets[gdb_bfd_section_index ((s)->objfile->obfd, (s)->the_bfd_section)])
/* The memory address of section S (vma + offset). */
#define obj_section_addr(s) \
(bfd_get_section_vma ((s)->objfile->obfd, s->the_bfd_section) \
+ obj_section_offset (s))
/* The one-passed-the-end memory address of section S
(vma + size + offset). */
#define obj_section_endaddr(s) \
(bfd_get_section_vma ((s)->objfile->obfd, s->the_bfd_section) \
+ bfd_get_section_size ((s)->the_bfd_section) \
+ obj_section_offset (s))
/* The "objstats" structure provides a place for gdb to record some
interesting information about its internal state at runtime, on a
per objfile basis, such as information about the number of symbols
read, size of string table (if any), etc. */
struct objstats
{
/* Number of partial symbols read. */
int n_psyms = 0;
/* Number of full symbols read. */
int n_syms = 0;
/* Number of ".stabs" read (if applicable). */
int n_stabs = 0;
/* Number of types. */
int n_types = 0;
/* Size of stringtable, (if applicable). */
int sz_strtab = 0;
};
#define OBJSTAT(objfile, expr) (objfile -> stats.expr)
#define OBJSTATS struct objstats stats
extern void print_objfile_statistics (void);
extern void print_symbol_bcache_statistics (void);
/* Number of entries in the minimal symbol hash table. */
#define MINIMAL_SYMBOL_HASH_SIZE 2039
/* Some objfile data is hung off the BFD. This enables sharing of the
data across all objfiles using the BFD. The data is stored in an
instance of this structure, and associated with the BFD using the
registry system. */
struct objfile_per_bfd_storage
{
objfile_per_bfd_storage ()
: minsyms_read (false)
{}
/* The storage has an obstack of its own. */
auto_obstack storage_obstack;
/* Byte cache for file names. */
bcache *filename_cache = NULL;
/* Byte cache for macros. */
bcache *macro_cache = NULL;
/* The gdbarch associated with the BFD. Note that this gdbarch is
determined solely from BFD information, without looking at target
information. The gdbarch determined from a running target may
differ from this e.g. with respect to register types and names. */
struct gdbarch *gdbarch = NULL;
/* Hash table for mapping symbol names to demangled names. Each
entry in the hash table is actually two consecutive strings,
both null-terminated; the first one is a mangled or linkage
name, and the second is the demangled name or just a zero byte
if the name doesn't demangle. */
htab *demangled_names_hash = NULL;
/* 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. */
entry_info ei {};
/* The name and language of any "main" found in this objfile. The
name can be NULL, which means that the information was not
recorded. */
const char *name_of_main = NULL;
enum language language_of_main = language_unknown;
/* 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
not include the terminating null symbol. The array itself, as well
as all the data that it points to, should be allocated on the
objfile_obstack for this file. */
minimal_symbol *msymbols = NULL;
int minimal_symbol_count = 0;
/* The number of minimal symbols read, before any minimal symbol
de-duplication is applied. Note in particular that this has only
a passing relationship with the actual size of the table above;
use minimal_symbol_count if you need the true size. */
int n_minsyms = 0;
/* This is true if minimal symbols have already been read. Symbol
readers can use this to bypass minimal symbol reading. Also, the
minimal symbol table management code in minsyms.c uses this to
suppress new minimal symbols. You might think that MSYMBOLS or
MINIMAL_SYMBOL_COUNT could be used for this, but it is possible
for multiple readers to install minimal symbols into a given
per-BFD. */
bool minsyms_read : 1;
/* This is a hash table used to index the minimal symbols by name. */
minimal_symbol *msymbol_hash[MINIMAL_SYMBOL_HASH_SIZE] {};
/* This hash table is used to index the minimal symbols by their
demangled names. */
minimal_symbol *msymbol_demangled_hash[MINIMAL_SYMBOL_HASH_SIZE] {};
/* All the different languages of symbols found in the demangled
hash table. A flat/vector-based map is more efficient than a map
or hash table here, since this will only usually contain zero or
one entries. */
std::vector demangled_hash_languages;
};
/* Master structure for keeping track of each file from which
gdb reads symbols. There are several ways these get allocated: 1.
The main symbol file, symfile_objfile, set by the symbol-file command,
2. Additional symbol files added by the add-symbol-file command,
3. Shared library objfiles, added by ADD_SOLIB, 4. symbol files
for modules that were loaded when GDB attached to a remote system
(see remote-vx.c). */
struct objfile
{
objfile (bfd *, const char *, objfile_flags);
~objfile ();
DISABLE_COPY_AND_ASSIGN (objfile);
/* All struct objfile's are chained together by their next pointers.
The program space field "objfiles" (frequently referenced via
the macro "object_files") points to the first link in this chain. */
struct objfile *next = nullptr;
/* The object file's original name as specified by the user,
made absolute, and tilde-expanded. However, it is not canonicalized
(i.e., it has not been passed through gdb_realpath).
This pointer is never NULL. This does not have to be freed; it is
guaranteed to have a lifetime at least as long as the objfile. */
char *original_name = nullptr;
CORE_ADDR addr_low = 0;
/* Some flag bits for this objfile. */
objfile_flags flags;
/* The program space associated with this objfile. */
struct program_space *pspace;
/* List of compunits.
These are used to do symbol lookups and file/line-number lookups. */
struct compunit_symtab *compunit_symtabs = nullptr;
/* 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 = nullptr;
/* Map addresses to the entries of PSYMTABS. It would be more efficient to
have a map per the whole process but ADDRMAP cannot selectively remove
its items during FREE_OBJFILE. This mapping is already present even for
PARTIAL_SYMTABs which still have no corresponding full SYMTABs read. */
struct addrmap *psymtabs_addrmap = nullptr;
/* List of freed partial symtabs, available for re-use. */
struct partial_symtab *free_psymtabs = nullptr;
/* The object file's BFD. Can be null if the objfile contains only
minimal symbols, e.g. the run time common symbols for SunOS4. */
bfd *obfd;
/* The per-BFD data. Note that this is treated specially if OBFD
is NULL. */
struct objfile_per_bfd_storage *per_bfd = nullptr;
/* The modification timestamp of the object file, as of the last time
we read its symbols. */
long mtime = 0;
/* Obstack to hold objects that should be freed when we load a new symbol
table from this object file. */
struct obstack objfile_obstack {};
/* A byte cache where we can stash arbitrary "chunks" of bytes that
will not change. */
struct psymbol_bcache *psymbol_cache;
/* Map symbol addresses to the partial symtab that defines the
object at that address. */
std::vector> psymbol_map;
/* Vectors of all partial symbols read in from file. The actual data
is stored in the objfile_obstack. */
std::vector global_psymbols;
std::vector static_psymbols;
/* 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. */
const struct sym_fns *sf = nullptr;
/* Per objfile data-pointers required by other GDB modules. */
REGISTRY_FIELDS {};
/* Set of relocation offsets to apply to each section.
The table is indexed by the_bfd_section->index, thus it is generally
as large as the number of sections in the binary.
The table is stored on the objfile_obstack.
These offsets indicate that all symbols (including partial and
minimal symbols) which have been read have been relocated by this
much. Symbols which are yet to be read need to be relocated by it. */
struct section_offsets *section_offsets = nullptr;
int num_sections = 0;
/* Indexes in the section_offsets array. These are initialized by the
*_symfile_offsets() family of functions (som_symfile_offsets,
xcoff_symfile_offsets, default_symfile_offsets). In theory they
should correspond to the section indexes used by bfd for the
current objfile. The exception to this for the time being is the
SOM version.
These are initialized to -1 so that we can later detect if they
are used w/o being properly assigned to. */
int sect_index_text = -1;
int sect_index_data = -1;
int sect_index_bss = -1;
int sect_index_rodata = -1;
/* These pointers are used to locate the section table, which
among other things, is used to map pc addresses into sections.
SECTIONS points to the first entry in the table, and
SECTIONS_END points to the first location past the last entry
in the table. The table is stored on the objfile_obstack. The
sections are indexed by the BFD section index; but the
structure data is only valid for certain sections
(e.g. non-empty, SEC_ALLOC). */
struct obj_section *sections = nullptr;
struct obj_section *sections_end = nullptr;
/* GDB allows to have debug symbols in separate object files. This is
used by .gnu_debuglink, ELF build id note and Mach-O OSO.
Although this is a tree structure, GDB only support one level
(ie a separate debug for a separate debug is not supported). Note that
separate debug object are in the main chain and therefore will be
visited by all_objfiles & co iterators. Separate debug objfile always
has a non-nul separate_debug_objfile_backlink. */
/* Link to the first separate debug object, if any. */
struct objfile *separate_debug_objfile = nullptr;
/* If this is a separate debug object, this is used as a link to the
actual executable objfile. */
struct objfile *separate_debug_objfile_backlink = nullptr;
/* If this is a separate debug object, this is a link to the next one
for the same executable objfile. */
struct objfile *separate_debug_objfile_link = nullptr;
/* Place to stash various statistics about this objfile. */
OBJSTATS;
/* A linked list of symbols created when reading template types or
function templates. These symbols are not stored in any symbol
table, so we have to keep them here to relocate them
properly. */
struct symbol *template_symbols = nullptr;
/* Associate a static link (struct dynamic_prop *) to all blocks (struct
block *) that have one.
In the context of nested functions (available in Pascal, Ada and GNU C,
for instance), a static link (as in DWARF's DW_AT_static_link attribute)
for a function is a way to get the frame corresponding to the enclosing
function.
Very few blocks have a static link, so it's more memory efficient to
store these here rather than in struct block. Static links must be
allocated on the objfile's obstack. */
htab_t static_links {};
};
/* Declarations for functions defined in objfiles.c */
extern struct gdbarch *get_objfile_arch (const struct objfile *);
extern int entry_point_address_query (CORE_ADDR *entry_p);
extern CORE_ADDR entry_point_address (void);
extern void build_objfile_section_table (struct objfile *);
extern struct objfile *objfile_separate_debug_iterate (const struct objfile *,
const struct objfile *);
extern void put_objfile_before (struct objfile *, struct objfile *);
extern void add_separate_debug_objfile (struct objfile *, struct objfile *);
extern void unlink_objfile (struct objfile *);
extern void free_objfile_separate_debug (struct objfile *);
extern void free_all_objfiles (void);
extern void objfile_relocate (struct objfile *, const struct section_offsets *);
extern void objfile_rebase (struct objfile *, CORE_ADDR);
extern int objfile_has_partial_symbols (struct objfile *objfile);
extern int objfile_has_full_symbols (struct objfile *objfile);
extern int objfile_has_symbols (struct objfile *objfile);
extern int have_partial_symbols (void);
extern int have_full_symbols (void);
extern void objfile_set_sym_fns (struct objfile *objfile,
const struct sym_fns *sf);
extern void objfiles_changed (void);
extern int is_addr_in_objfile (CORE_ADDR addr, const struct objfile *objfile);
/* Return true if ADDRESS maps into one of the sections of a
OBJF_SHARED objfile of PSPACE and false otherwise. */
extern int shared_objfile_contains_address_p (struct program_space *pspace,
CORE_ADDR address);
/* This operation deletes all objfile entries that represent solibs that
weren't explicitly loaded by the user, via e.g., the add-symbol-file
command. */
extern void objfile_purge_solibs (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 (void);
extern struct obj_section *find_pc_section (CORE_ADDR pc);
/* Return non-zero if PC is in a section called NAME. */
extern int pc_in_section (CORE_ADDR, const char *);
/* Return non-zero if PC is in a SVR4-style procedure linkage table
section. */
static inline int
in_plt_section (CORE_ADDR pc)
{
return pc_in_section (pc, ".plt");
}
/* Keep a registry of per-objfile data-pointers required by other GDB
modules. */
DECLARE_REGISTRY(objfile);
/* In normal use, the section map will be rebuilt by find_pc_section
if objfiles have been added, removed or relocated since it was last
called. Calling inhibit_section_map_updates will inhibit this
behavior until the returned scoped_restore object is destroyed. If
you call inhibit_section_map_updates you must ensure that every
call to find_pc_section in the inhibited region relates to a
section that is already in the section map and has not since been
removed or relocated. */
extern scoped_restore_tmpl inhibit_section_map_updates
(struct program_space *pspace);
extern void default_iterate_over_objfiles_in_search_order
(struct gdbarch *gdbarch,
iterate_over_objfiles_in_search_order_cb_ftype *cb,
void *cb_data, struct objfile *current_objfile);
/* An iterarable object that can be used to iterate over all
objfiles. The basic use is in a foreach, like:
for (objfile *objf : all_objfiles (pspace)) { ... } */
class all_objfiles : public next_adapter
{
public:
explicit all_objfiles (struct program_space *pspace)
: next_adapter (pspace->objfiles)
{
}
};
/* An iterarable object that can be used to iterate over all
objfiles. The basic use is in a foreach, like:
for (objfile *objf : all_objfiles_safe (pspace)) { ... }
This variant uses a basic_safe_iterator so that objfiles can be
deleted during iteration. */
class all_objfiles_safe
: public next_adapter>>
{
public:
explicit all_objfiles_safe (struct program_space *pspace)
: next_adapter>>
(pspace->objfiles)
{
}
};
/* A range adapter that makes it possible to iterate over all
compunits in one objfile. */
class objfile_compunits : public next_adapter
{
public:
explicit objfile_compunits (struct objfile *objfile)
: next_adapter (objfile->compunit_symtabs)
{
}
};
/* A range adapter that makes it possible to iterate over all
minimal symbols of an objfile. */
class objfile_msymbols
{
public:
explicit objfile_msymbols (struct objfile *objfile)
: m_objfile (objfile)
{
}
struct iterator
{
typedef iterator self_type;
typedef struct minimal_symbol *value_type;
typedef struct minimal_symbol *&reference;
typedef struct minimal_symbol **pointer;
typedef std::forward_iterator_tag iterator_category;
typedef int difference_type;
explicit iterator (struct objfile *objfile)
: m_msym (objfile->per_bfd->msymbols)
{
/* Make sure to properly handle the case where there are no
minsyms. */
if (MSYMBOL_LINKAGE_NAME (m_msym) == nullptr)
m_msym = nullptr;
}
iterator ()
: m_msym (nullptr)
{
}
value_type operator* () const
{
return m_msym;
}
bool operator== (const self_type &other) const
{
return m_msym == other.m_msym;
}
bool operator!= (const self_type &other) const
{
return m_msym != other.m_msym;
}
self_type &operator++ ()
{
if (m_msym != nullptr)
{
++m_msym;
if (MSYMBOL_LINKAGE_NAME (m_msym) == nullptr)
m_msym = nullptr;
}
return *this;
}
private:
struct minimal_symbol *m_msym;
};
iterator begin () const
{
return iterator (m_objfile);
}
iterator end () const
{
return iterator ();
}
private:
struct objfile *m_objfile;
};
#define ALL_OBJFILE_OSECTIONS(objfile, osect) \
for (osect = objfile->sections; osect < objfile->sections_end; osect++) \
if (osect->the_bfd_section == NULL) \
{ \
/* Nothing. */ \
} \
else
/* Traverse all obj_sections in all objfiles in the current program
space.
Note that this detects a "break" in the inner loop, and exits
immediately from the outer loop as well, thus, client code doesn't
need to know that this is implemented with a double for. The extra
hair is to make sure that a "break;" stops the outer loop iterating
as well, and both OBJFILE and OSECT are left unmodified:
- The outer loop learns about the inner loop's end condition, and
stops iterating if it detects the inner loop didn't reach its
end. In other words, the outer loop keeps going only if the
inner loop reached its end cleanly [(osect) ==
(objfile)->sections_end].
- OSECT is initialized in the outer loop initialization
expressions, such as if the inner loop has reached its end, so
the check mentioned above succeeds the first time.
- The trick to not clearing OBJFILE on a "break;" is, in the outer
loop's loop expression, advance OBJFILE, but iff the inner loop
reached its end. If not, there was a "break;", so leave OBJFILE
as is; the outer loop's conditional will break immediately as
well (as OSECT will be different from OBJFILE->sections_end). */
#define ALL_OBJSECTIONS(objfile, osect) \
for ((objfile) = current_program_space->objfiles, \
(objfile) != NULL ? ((osect) = (objfile)->sections_end) : 0; \
(objfile) != NULL \
&& (osect) == (objfile)->sections_end; \
((osect) == (objfile)->sections_end \
? ((objfile) = (objfile)->next, \
(objfile) != NULL ? (osect) = (objfile)->sections_end : 0) \
: 0)) \
ALL_OBJFILE_OSECTIONS (objfile, osect)
#define SECT_OFF_DATA(objfile) \
((objfile->sect_index_data == -1) \
? (internal_error (__FILE__, __LINE__, \
_("sect_index_data not initialized")), -1) \
: objfile->sect_index_data)
#define SECT_OFF_RODATA(objfile) \
((objfile->sect_index_rodata == -1) \
? (internal_error (__FILE__, __LINE__, \
_("sect_index_rodata not initialized")), -1) \
: objfile->sect_index_rodata)
#define SECT_OFF_TEXT(objfile) \
((objfile->sect_index_text == -1) \
? (internal_error (__FILE__, __LINE__, \
_("sect_index_text not initialized")), -1) \
: objfile->sect_index_text)
/* Sometimes the .bss section is missing from the objfile, so we don't
want to die here. Let the users of SECT_OFF_BSS deal with an
uninitialized section index. */
#define SECT_OFF_BSS(objfile) (objfile)->sect_index_bss
/* Answer whether there is more than one object file loaded. */
#define MULTI_OBJFILE_P() (object_files && object_files->next)
/* Reset the per-BFD storage area on OBJ. */
void set_objfile_per_bfd (struct objfile *obj);
/* Return canonical name for OBJFILE.
This is the real file name if the file has been opened.
Otherwise it is the original name supplied by the user. */
const char *objfile_name (const struct objfile *objfile);
/* Return the (real) file name of OBJFILE if the file has been opened,
otherwise return NULL. */
const char *objfile_filename (const struct objfile *objfile);
/* Return the name to print for OBJFILE in debugging messages. */
extern const char *objfile_debug_name (const struct objfile *objfile);
/* Return the name of the file format of OBJFILE if the file has been opened,
otherwise return NULL. */
const char *objfile_flavour_name (struct objfile *objfile);
/* Set the objfile's notion of the "main" name and language. */
extern void set_objfile_main_name (struct objfile *objfile,
const char *name, enum language lang);
extern void objfile_register_static_link
(struct objfile *objfile,
const struct block *block,
const struct dynamic_prop *static_link);
extern const struct dynamic_prop *objfile_lookup_static_link
(struct objfile *objfile, const struct block *block);
#endif /* !defined (OBJFILES_H) */