GDB

From Gentoo Wiki
(Redirected from Debugging with GDB)
Jump to:navigation Jump to:search

The GNU debugger (GDB), is used to find runtime errors that normally involve memory corruption.

Installation

USE flags

USE flags for dev-debug/gdb GNU debugger

cet Enable Intel Control-flow Enforcement Technology.
debuginfod Enable debuginfod support via dev-libs/elfutils libdebuginfod
guile Add support for the guile Scheme interpreter
lzma Support lzma compression in ELF debug info
multitarget Support all known targets in one gdb binary
nls Add Native Language Support (using gettext - GNU locale utilities)
python Enable support for the new internal scripting language, as well as extended pretty printers
server Install the "gdbserver" program (useful for embedded/remote targets)
sim Build gdb's simulators for various hardware platforms. See https://sourceware.org/gdb/wiki/Sim.
source-highlight Enable listing highlighting via dev-util/source-highlight
test Enable dependencies and/or preparations necessary to run tests (usually controlled by FEATURES=test but can be toggled independently)
vanilla Do not add extra patches which change default behaviour; DO NOT USE THIS ON A GLOBAL SCALE as the severity of the meaning changes drastically
xml Support parsing XML data files needed (at least) for cpu features, memory maps, and syscall tracing
xxhash Use dev-libs/xxhash to speed up internal hashing.
zstd Enable support for ZSTD compression

Emerge

root #emerge --ask dev-debug/gdb

Alternative interfaces

gdb is a complex program with many alternative interfaces, some optimized for different use cases:

Configuration

Files

gdb can be configured via the following files:

  • ~/.config/gdb/gdbearlyinit which is checked before any other configuration file
  • ~/.config/gdb/gdbinit which is the primary configuration file

Tweaks

Silent startup

By default, gdb is quite noisy at startup:

user $gdb
GNU gdb (Gentoo 13.0.90_p20230110 vanilla) 13.0.90.20230110-git
Copyright (C) 2023 Free Software Foundation, Inc.
License GPLv3+: GNU GPL version 3 or later <http://gnu.org/licenses/gpl.html>
This is free software: you are free to change and redistribute it.
There is NO WARRANTY, to the extent permitted by law.
Type "show copying" and "show warranty" for details.
This GDB was configured as "x86_64-pc-linux-gnu".
Type "show configuration" for configuration details.
For bug reporting instructions, please see:
<https://bugs.gentoo.org/>.
Find the GDB manual and other documentation resources online at:
    <http://www.gnu.org/software/gdb/documentation/>.

For help, type "help".
Type "apropos word" to search for commands related to "word".
(gdb)

To change this, enable startup-quietly:

FILE ~/.config/gdb/gdbearlyinit
set startup-quietly on

Then afterwards, gdb is far quieter:

user $gdb
(gdb)

Persistent history

By default, gdb does not retain command history or scrollback between runs. This can be changed via the various history toggles.

FILE ~/.config/gdb/gdbinit
set history filename ~/.gdb_history
set history save on
set history size unlimited
set history remove-duplicates unlimited

Retain debug symbols

See Installing debugging information for packages.

In order to debug a program, emerge the program either with FEATURES="nostrip" to prevent the stripping of debug symbols, or with the needed package.env settings.

Once debug symbols are set, it is time to begin debugging the program.

If more assistance is needed, such as being able to see the code within GDB TUI (GDB Text User Interface), install the dev-util/debugedit package and include FEATURES="installsources".

Although not really needed for meaningful backtraces, this significantly helps when actively coding.

Usage

Enabling core dumps

Core dumps can also be used for debugging. These core files (named for when RAM was called "core" memory) contain the same information that the program would produce when run with gdb. Here, bad_code is the program that will be debugged with gdb. In order to debug with a core file with bad_code, run gdb ./bad_code core where core is the name of the core file.

Sometimes a program crashes without an easy way to run it manually under gdb. A solution to this can be obtaining a coredump.

If the program's output is as follows, there was no coredump created:

user $./foo
Segmentation fault

Note that it doesn't say "(core dumped)". If coredumps are already enabled, the following output will appear:

user $./foo
Segmentation fault (core dumped)

To enable core dumps, run the following in a shell:

root #sysctl kernel.yama.ptrace_scope=0 && echo 2 > /proc/sys/fs/suid_dumpable # Allows running as a normal user
root #ulimit -c unlimited # Allow any size for coredumps (re-run this as the user running the broken program if needed)

Then running the program again, a corefile should be created:

user $./foo
Segmentation fault (core dumped)
user $gdb foo foo.core
Reading symbols from foo...
Core was generated by `./foo'.
#0  0x00007f4fc7b37ae0 in gets () from /usr/lib64/libc.so.6
(gdb) bt
#0  0x00007f4fc7b37ae0 in gets () from /usr/lib64/libc.so.6
#1  0x000055d73f62e08f in main () at /tmp/foo.c:10
(gdb)

Note that Systemd users can utilize coredumpctl as well, while OpenRC users may wish to try sys-process/minicoredumper.

Running the program with GDB

Tip
Consider building the program with additional warnings. -Wall prints all compiler warnings. See GCC Debugging Options for details.

Suppose there is a program which needs to be debugged called bad_code. Some person claims that the program crashes and provides an example. Go ahead and test it out:

user $./bad_code `perl -e 'print "A"x100'`
Segmentation fault

It seems this person was right. Since the program is obviously broken, there is now a bug at hand. Now, it is time to use gdb to help solve this matter. First, run gdb with the --args option, then give it the full program with arguments:

user $gdb --args ./bad_code `perl -e 'print "A"x100'`
GNU gdb 6.3
Copyright 2004 Free Software Foundation, Inc.
GDB is free software, covered by the GNU General Public License, and you are
welcome to change it and/or distribute copies of it under certain conditions.
Type "show copying" to see the conditions.
There is absolutely no warranty for GDB. Type "show warranty" for details.
This GDB was configured as "i686-pc-linux-gnu"...Using host libthread_db library "/lib/libthread_db.so.1".

A prompt that says "(gdb)" should display and wait for input. First, run the program with run:

(gdb)run
Starting program: /home/chris/bad_code

Program received signal SIGSEGV, Segmentation fault.
0xb7ec6dc0 in strcpy () from /lib/libc.so.6

The program should start, and show a notification of SIGSEGV, or Segmentation Fault. This is GDB telling us that the program has crashed. It also gives the last run function it could trace when the program crashes. However, this is not too useful, as there could be multiple strcpy's in the program, making it hard for developers to find which one is causing the issue. In order to help them out, a backtrace can be performed.

A backtrace runs backwards through all the functions that occurred upon program execution, to the function at fault. Functions that return (without causing a crash) will not show up on the backtrace. To get a backtrace at the (gdb) prompt type in bt. Something similar to the following message should be returned:

(gdb)bt
#0  0xb7ec6dc0 in strcpy () from /lib/libc.so.6
#1  0x0804838c in run_it ()
#2  0x080483ba in main ()

The trace pattern should be clearly visible. main() is called first, followed by run_it(), and somewhere in run_it() lies the strcpy() at fault. Things such as this help developers narrow down problems. There are a few exceptions to the output. First off is forgetting to enable debug symbols with FEATURES="nostrip". With debug symbols stripped, the output looks something like this:

(gdb)bt
#0  0xb7e2cdc0 in strcpy () from /lib/libc.so.6
#1  0x0804838c in ?? ()
#2  0xbfd19510 in ?? ()
#3  0x00000000 in ?? ()
#4  0x00000000 in ?? ()
#5  0xb7eef148 in libgcc_s_personality () from /lib/libc.so.6
#6  0x080482ed in ?? ()
#7  0x080495b0 in ?? ()
#8  0xbfd19528 in ?? ()
#9  0xb7dd73b8 in __guard_setup () from /lib/libc.so.6
#10 0xb7dd742d in __guard_setup () from /lib/libc.so.6
#11 0x00000006 in ?? ()
#12 0xbfd19548 in ?? ()
#13 0x080483ba in ?? ()
#14 0x00000000 in ?? ()
#15 0x00000000 in ?? ()
#16 0xb7deebcc in __new_exitfn () from /lib/libc.so.6
#17 0x00000000 in ?? ()
#18 0xbfd19560 in ?? ()
#19 0xb7ef017c in nullserv () from /lib/libc.so.6
#20 0xb7dd6f37 in __libc_start_main () from /lib/libc.so.6
#21 0x00000001 in ?? ()
#22 0xbfd195d4 in ?? ()
#23 0xbfd195dc in ?? ()
#24 0x08048201 in ?? ()

This backtrace contains a large number of ?? (question) marks. This is because without debug symbols, gdb does not know how the program was run. Hence, it is crucial that debug symbols are not stripped.

Here is what the output looks like with -ggdb3 flag enabled:

(gdb)bt
#0  0xb7e4bdc0 in strcpy () from /lib/libc.so.6
#1  0x0804838c in run_it (input=0x0) at bad_code.c:7
#2  0x080483ba in main (argc=1, argv=0xbfd3a434) at bad_code.c:12

Here, a lot more information is available for developers. Not only is function information displayed, but even the exact line numbers of the source files. This method is the most preferred, as long as the extra space can be spared.

The backtrace can be saved to a file by copying and pasting from the terminal. Once gdb tasks are over, quit.

(gdb)quit
The program is running. Exit anyway? (y or n) y

This ends the walk-through of gdb. However, there are other types of errors such as improper file access that can cause a program to fail during runtime. Find those using strace.

Enabling ASLR

gdb disables ASLR by default to make addresses more predictable, but this can affect how reproducible a bug is. It can mean a bug doesn't occur within gdb in some rare circumstances.

If this happens, try set disable-randomization off within gdb before running the program with r.

See also