binutils-gdb/gdb/gdbcore.h
Philippe Waroquiers a2fedca99c Implement 'set/show exec-file-mismatch'.
This option allows to tell GDB to detect and possibly handle mismatched exec-files.

A recurrent problem with GDB is that GDB uses the wrong exec-file
when using the attach/detach commands successively.
Also, in case the user specifies a file on the command line but attaches
to the wrong PID, this error is not made visible and gives a not user
understandable behaviour.

For example:
  $ gdb
  ...
  (gdb) atta 2682  ############################################  PID running 'sleepers' executable
  Attaching to process 2682
  [New LWP 2683]
  [New LWP 2684]
  [New LWP 2685]
  [Thread debugging using libthread_db enabled]
  Using host libthread_db library "/lib/x86_64-linux-gnu/libthread_db.so.1".
  0x00007f5ff829f603 in select () at ../sysdeps/unix/syscall-template.S:84
  84    ../sysdeps/unix/syscall-template.S: No such file or directory.
  (gdb) det
  Detaching from program: /home/philippe/valgrind/git/trunk_untouched/gdbserver_tests/sleepers, process 2682
  [Inferior 1 (process 2682) detached]
  (gdb) atta 31069 ############################################  PID running 'gdb' executable
  Attaching to program: /home/philippe/valgrind/git/trunk_untouched/gdbserver_tests/sleepers, process 31069
  Reading symbols from /lib64/ld-linux-x86-64.so.2...
  Reading symbols from /usr/lib/debug/.build-id/60/6df9c355103e82140d513bc7a25a635591c153.debug...
  0x00007f43c23478a0 in ?? ()
  (gdb) bt
  #0  0x00007f43c23478a0 in ?? ()
  #1  0x0000558909e3ad91 in ?? ()
  #2  0x0000202962646700 in ?? ()
  #3  0x00007ffc69c74e70 in ?? ()
  #4  0x000055890c1d2350 in ?? ()
  #5  0x0000000000000000 in ?? ()
  (gdb)

The second attach has kept the executable of the first attach.
(in this case, 31069 is the PID of a GDB, that has nothing to do
with the first determined 'sleepers' executable).

Similarly, if specifying an executable, but attaching to a wrong pid,
we get:

  gdb /home/philippe/valgrind/git/trunk_untouched/gdbserver_tests/sleepers
  ...
  Reading symbols from /home/philippe/valgrind/git/trunk_untouched/gdbserver_tests/sleepers...
  (gdb) atta 31069 ############################################  PID running 'gdb' executable
  Attaching to program: /home/philippe/valgrind/git/trunk_untouched/gdbserver_tests/sleepers, process 31069
  Reading symbols from /lib64/ld-linux-x86-64.so.2...
  Reading symbols from /usr/lib/debug/.build-id/60/6df9c355103e82140d513bc7a25a635591c153.debug...
  0x00007f43c23478a0 in ?? ()
  (gdb) bt
  #0  0x00007f43c23478a0 in ?? ()
  #1  0x0000558909e3ad91 in ?? ()
  #2  0x0000202962646700 in ?? ()
  #3  0x00007ffc69c74e70 in ?? ()
  #4  0x000055890c1d2350 in ?? ()
  #5  0x0000000000000000 in ?? ()
  (gdb)

And it is unclear to the user what has happened/what is going wrong.

This patch series implements a new option:
    (gdb) apropos exec-file-mismatch
    set exec-file-mismatch -- Set exec-file-mismatch handling (ask|warn|off).
    show exec-file-mismatch -- Show exec-file-mismatch handling (ask|warn|off).
    (gdb) help set exec-file-mismatch
    Set exec-file-mismatch handling (ask|warn|off).
    Specifies how to handle a mismatch between the current exec-file name
    loaded by GDB and the exec-file name automatically determined when attaching
    to a process:

     ask  - warn the user and ask whether to load the determined exec-file.
     warn - warn the user, but do not change the exec-file.
     off  - do not check for mismatch.

"ask" means: in case of mismatch between the current exec-file name
and the automatically determined exec-file name of the PID we are attaching to,
give a warning to the user and ask whether to load the automatically determined
exec-file.

"warn" means: in case of mismatch, just give a warning to the user.

"off" means: do not check for mismatch.

This fixes PR gdb/17626.
There was a previous trial to fix this PR.
See https://sourceware.org/ml/gdb-patches/2015-07/msg00118.html
This trial was however only fixing the problem for the automatically
determined executable files when doing attach.
It was differentiating the 'user specified executable files' ("sticky")
from the executable files automatically found by GDB.
But such user specified sticky executables are in most cases due
to a wrong manipulation by the user, giving unexpected results
such as backtrace showing no function like in the above example.

This patch ensures that whenever a process executable can be
determined, that the user is warned if there is a mismatch.

The same tests as above then give:

  (gdb) atta 2682
  Attaching to process 2682
  [New LWP 2683]
  [New LWP 2684]
  [New LWP 2685]
  [Thread debugging using libthread_db enabled]
  Using host libthread_db library "/lib/x86_64-linux-gnu/libthread_db.so.1".
  0x00007f5ff829f603 in select () at ../sysdeps/unix/syscall-template.S:84
  84    ../sysdeps/unix/syscall-template.S: No such file or directory.
  (gdb) det
  Detaching from program: /home/philippe/valgrind/git/trunk_untouched/gdbserver_tests/sleepers, process 2682
  [Inferior 1 (process 2682) detached]
  (gdb) atta 31069
  Attaching to program: /home/philippe/valgrind/git/trunk_untouched/gdbserver_tests/sleepers, process 31069
  warning: Mismatch between current exec-file /home/philippe/valgrind/git/trunk_untouched/gdbserver_tests/sleepers
  and automatically determined exec-file /bd/home/philippe/gdb/git/build_fixes/gdb/gdb
  exec-file-mismatch handling is currently "ask"
  Load new symbol table from "/bd/home/philippe/gdb/git/build_fixes/gdb/gdb"? (y or n) y
  Reading symbols from /bd/home/philippe/gdb/git/build_fixes/gdb/gdb...
  Setting up the environment for debugging gdb.
  ...
  Reading symbols from /usr/lib/debug/.build-id/60/6df9c355103e82140d513bc7a25a635591c153.debug...
  0x00007f43c23478a0 in __poll_nocancel () at ../sysdeps/unix/syscall-template.S:84
  84    ../sysdeps/unix/syscall-template.S: No such file or directory.
  (top-gdb) bt
  During symbol reading: incomplete CFI data; unspecified registers (e.g., rax) at 0x7f43c23478ad
  During symbol reading: unsupported tag: 'DW_TAG_unspecified_type'
  During symbol reading: cannot get low and high bounds for subprogram DIE at 0x12282a7
  During symbol reading: Child DIE 0x12288ba and its abstract origin 0x1228b26 have different parents
  During symbol reading: DW_AT_call_target target DIE has invalid low pc, for referencing DIE 0x1229540 [in module /bd/home/philippe/gdb/git/build_fixes/gdb/gdb]
  #0  0x00007f43c23478a0 in __poll_nocancel () at ../sysdeps/unix/syscall-template.S:84
  #1  0x0000558909e3ad91 in poll (__timeout=-1, __nfds=<optimized out>, __fds=<optimized out>) at /usr/include/x86_64-linux-gnu/bits/poll2.h:46
  #2  gdb_wait_for_event (block=block@entry=1) at ../../fixes/gdb/event-loop.c:772
  #3  0x0000558909e3aef4 in gdb_do_one_event () at ../../fixes/gdb/event-loop.c:347
  #4  0x0000558909e3b085 in gdb_do_one_event () at ../../fixes/gdb/common/common-exceptions.h:219
  #5  start_event_loop () at ../../fixes/gdb/event-loop.c:371
  During symbol reading: Member function "~_Sp_counted_base" (offset 0x1c69bf7) is virtual but the vtable offset is not specified
  During symbol reading: Multiple children of DIE 0x1c8f5a0 refer to DIE 0x1c8f0ee as their abstract origin
  #6  0x0000558909ed3b78 in captured_command_loop () at ../../fixes/gdb/main.c:331
  #7  0x0000558909ed4b6d in captured_main (data=<optimized out>) at ../../fixes/gdb/main.c:1174
  #8  gdb_main (args=<optimized out>) at ../../fixes/gdb/main.c:1190
  #9  0x0000558909c1e9a8 in main (argc=<optimized out>, argv=<optimized out>) at ../../fixes/gdb/gdb.c:32
  (top-gdb)

  gdb /home/philippe/valgrind/git/trunk_untouched/gdbserver_tests/sleepers
  ...
  Reading symbols from /home/philippe/valgrind/git/trunk_untouched/gdbserver_tests/sleepers...
  (gdb) atta 31069
  Attaching to program: /home/philippe/valgrind/git/trunk_untouched/gdbserver_tests/sleepers, process 31069
  warning: Mismatch between current exec-file /home/philippe/valgrind/git/trunk_untouched/gdbserver_tests/sleepers
  and automatically determined exec-file /bd/home/philippe/gdb/git/build_fixes/gdb/gdb
  exec-file-mismatch handling is currently "ask"
  Load new symbol table from "/bd/home/philippe/gdb/git/build_fixes/gdb/gdb"? (y or n) y
  Reading symbols from /bd/home/philippe/gdb/git/build_fixes/gdb/gdb...
  Setting up the environment for debugging gdb.
  ....

In other words, it now works as intuitively expected by the user.
If ever the user gave the correct executable on the command line,
then attached to the wrong pid, then confirmed loading the wrong executable,
the user can simply fix this by detaching, and attaching to the correct pid,
GDB will then tell again to the user that the exec-file might better
be loaded.

The default value of "ask" is chosen instead of e.g. "warn" as in most
cases, switching of executable will be the correct action,
and in any case, the user can decide to not load the executable,
as GDB asks a confirmation to the user to load the new executable.

For settings "ask" and "warn", the new function validate_exec_file ()
tries to get the inferior pid exec file and compares it with the current
exec file.  In case of mismatch, it warns the user and optionally load
the executable.
This function is called in the attach_command implementation to cover
most cases of attaching to a running process.
It must also be called in remote.c, as the attach command is not supported
for all types of remote gdbserver.

gdb/ChangeLog
2020-01-25  Philippe Waroquiers  <philippe.waroquiers@skynet.be>

	* exec.c (exec_file_mismatch_names, exec_file_mismatch_mode)
	(show_exec_file_mismatch_command, set_exec_file_mismatch_command)
	(validate_exec_file): New variables, enums, functions.
	(exec_file_locate_attach, print_section_info): Style the filenames.
	(_initialize_exec): Install show_exec_file_mismatch_command and
	 set_exec_file_mismatch_command.
	* gdbcore.h (validate_exec_file): Declare.
	* infcmd.c (attach_command): Call validate_exec_file.
	* remote.c ( remote_target::remote_add_inferior): Likewise.
2020-01-25 11:05:59 +01:00

283 lines
9.6 KiB
C++
Raw Blame History

This file contains invisible Unicode characters

This file contains invisible Unicode characters that are indistinguishable to humans but may be processed differently by a computer. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.

/* Machine independent variables that describe the core file under GDB.
Copyright (C) 1986-2020 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 <http://www.gnu.org/licenses/>. */
/* Interface routines for core, executable, etc. */
#if !defined (GDBCORE_H)
#define GDBCORE_H 1
struct type;
struct regcache;
#include "bfd.h"
#include "exec.h"
#include "target.h"
/* Nonzero if there is a core file. */
extern int have_core_file_p (void);
/* Report a memory error with error(). */
extern void memory_error (enum target_xfer_status status, CORE_ADDR memaddr);
/* The string 'memory_error' would use as exception message. */
extern std::string memory_error_message (enum target_xfer_status err,
struct gdbarch *gdbarch,
CORE_ADDR memaddr);
/* Like target_read_memory, but report an error if can't read. */
extern void read_memory (CORE_ADDR memaddr, gdb_byte *myaddr, ssize_t len);
/* Like target_read_stack, but report an error if can't read. */
extern void read_stack (CORE_ADDR memaddr, gdb_byte *myaddr, ssize_t len);
/* Like target_read_code, but report an error if can't read. */
extern void read_code (CORE_ADDR memaddr, gdb_byte *myaddr, ssize_t len);
/* Read an integer from debugged memory, given address and number of
bytes. */
extern LONGEST read_memory_integer (CORE_ADDR memaddr,
int len, enum bfd_endian byte_order);
extern int safe_read_memory_integer (CORE_ADDR memaddr, int len,
enum bfd_endian byte_order,
LONGEST *return_value);
/* Read an unsigned integer from debugged memory, given address and
number of bytes. */
extern ULONGEST read_memory_unsigned_integer (CORE_ADDR memaddr,
int len,
enum bfd_endian byte_order);
extern int safe_read_memory_unsigned_integer (CORE_ADDR memaddr, int len,
enum bfd_endian byte_order,
ULONGEST *return_value);
/* Read an integer from debugged code memory, given address,
number of bytes, and byte order for code. */
extern LONGEST read_code_integer (CORE_ADDR memaddr, int len,
enum bfd_endian byte_order);
/* Read an unsigned integer from debugged code memory, given address,
number of bytes, and byte order for code. */
extern ULONGEST read_code_unsigned_integer (CORE_ADDR memaddr,
int len,
enum bfd_endian byte_order);
/* Read a null-terminated string from the debuggee's memory, given
address, a buffer into which to place the string, and the maximum
available space. */
extern void read_memory_string (CORE_ADDR, char *, int);
/* Read the pointer of type TYPE at ADDR, and return the address it
represents. */
CORE_ADDR read_memory_typed_address (CORE_ADDR addr, struct type *type);
/* Same as target_write_memory, but report an error if can't
write. */
extern void write_memory (CORE_ADDR memaddr, const gdb_byte *myaddr,
ssize_t len);
/* Same as write_memory, but notify 'memory_changed' observers. */
extern void write_memory_with_notification (CORE_ADDR memaddr,
const bfd_byte *myaddr,
ssize_t len);
/* Store VALUE at ADDR in the inferior as a LEN-byte unsigned integer. */
extern void write_memory_unsigned_integer (CORE_ADDR addr, int len,
enum bfd_endian byte_order,
ULONGEST value);
/* Store VALUE at ADDR in the inferior as a LEN-byte unsigned integer. */
extern void write_memory_signed_integer (CORE_ADDR addr, int len,
enum bfd_endian byte_order,
LONGEST value);
/* Hook for `exec_file_command' command to call. */
extern void (*deprecated_exec_file_display_hook) (const char *filename);
/* Hook for "file_command", which is more useful than above
(because it is invoked AFTER symbols are read, not before). */
extern void (*deprecated_file_changed_hook) (const char *filename);
extern void specify_exec_file_hook (void (*hook) (const char *filename));
/* Binary File Diddler for the core file. */
#define core_bfd (current_program_space->cbfd.get ())
/* Whether to open exec and core files read-only or read-write. */
extern bool write_files;
/* Open and set up the core file bfd. */
extern void core_target_open (const char *arg, int from_tty);
extern void core_file_command (const char *filename, int from_tty);
extern void exec_file_attach (const char *filename, int from_tty);
/* If the filename of the main executable is unknown, attempt to
determine it. If a filename is determined, proceed as though
it was just specified with the "file" command. Do nothing if
the filename of the main executable is already known.
DEFER_BP_RESET uses SYMFILE_DEFER_BP_RESET for the main symbol file. */
extern void exec_file_locate_attach (int pid, int defer_bp_reset, int from_tty);
extern void validate_files (void);
/* Give the user a message if the current exec file does not match the exec
file determined from the target. In case of mismatch, ask the user
if the exec file determined from target must be loaded. */
extern void validate_exec_file (int from_tty);
/* The current default bfd target. */
extern char *gnutarget;
extern void set_gnutarget (const char *);
/* Structure to keep track of core register reading functions for
various core file types. */
struct core_fns
{
/* BFD flavour that a core file handler is prepared to read. This
can be used by the handler's core tasting function as a first
level filter to reject BFD's that don't have the right
flavour. */
enum bfd_flavour core_flavour;
/* Core file handler function to call to recognize corefile
formats that BFD rejects. Some core file format just don't fit
into the BFD model, or may require other resources to identify
them, that simply aren't available to BFD (such as symbols from
another file). Returns nonzero if the handler recognizes the
format, zero otherwise. */
int (*check_format) (bfd *);
/* Core file handler function to call to ask if it can handle a
given core file format or not. Returns zero if it can't,
nonzero otherwise. */
int (*core_sniffer) (struct core_fns *, bfd *);
/* Extract the register values out of the core file and supply them
into REGCACHE.
CORE_REG_SECT points to the register values themselves, read into
memory.
CORE_REG_SIZE is the size of that area.
WHICH says which set of registers we are handling:
0 --- integer registers
2 --- floating-point registers, on machines where they are
discontiguous
3 --- extended floating-point registers, on machines where
these are present in yet a third area. (GNU/Linux uses
this to get at the SSE registers.)
REG_ADDR is the offset from u.u_ar0 to the register values relative to
core_reg_sect. This is used with old-fashioned core files to locate the
registers in a large upage-plus-stack ".reg" section. Original upage
address X is at location core_reg_sect+x+reg_addr. */
void (*core_read_registers) (struct regcache *regcache,
gdb_byte *core_reg_sect,
unsigned core_reg_size,
int which, CORE_ADDR reg_addr);
/* Finds the next struct core_fns. They are allocated and
initialized in whatever module implements the functions pointed
to; an initializer calls deprecated_add_core_fns to add them to
the global chain. */
struct core_fns *next;
};
/* Build either a single-thread or multi-threaded section name for
PTID.
If ptid's lwp member is zero, we want to do the single-threaded
thing: look for a section named NAME (as passed to the
constructor). If ptid's lwp member is non-zero, we'll want do the
multi-threaded thing: look for a section named "NAME/LWP", where
LWP is the shortest ASCII decimal representation of ptid's lwp
member. */
class thread_section_name
{
public:
/* NAME is the single-threaded section name. If PTID represents an
LWP, then the build section name is "NAME/LWP", otherwise it's
just "NAME" unmodified. */
thread_section_name (const char *name, ptid_t ptid)
{
if (ptid.lwp_p ())
{
m_storage = string_printf ("%s/%ld", name, ptid.lwp ());
m_section_name = m_storage.c_str ();
}
else
m_section_name = name;
}
/* Return the computed section name. The result is valid as long as
this thread_section_name object is live. */
const char *c_str () const
{ return m_section_name; }
DISABLE_COPY_AND_ASSIGN (thread_section_name);
private:
/* Either a pointer into M_STORAGE, or a pointer to the name passed
as parameter to the constructor. */
const char *m_section_name;
/* If we need to build a new section name, this is where we store
it. */
std::string m_storage;
};
/* Replaced by the "iterate_over_regset_sections" gdbarch method. */
extern void deprecated_add_core_fns (struct core_fns *cf);
extern int default_core_sniffer (struct core_fns *cf, bfd * abfd);
extern int default_check_format (bfd * abfd);
#endif /* !defined (GDBCORE_H) */