nuttx/Documentation/quickstart/debugging.rst
anjiahao 4745ab64c3 Documentation:add gdb python script documentation
Signed-off-by: anjiahao <anjiahao@xiaomi.com>
2023-07-24 08:05:58 -03:00

271 lines
11 KiB
ReStructuredText

.. include:: /substitutions.rst
.. _debugging:
=========
Debugging
=========
Finding and fixing bugs is an important part of the hardware and software development process. Sometimes you also need
to use debugging techniques to understand how the system works. Two tools that are helpful are debug logging and
debugging using the GNU Debugger (gdb).
Debug Logging
=============
NuttX has a powerful system logging facility (syslog) with ``info``, ``warn``, and ``error`` levels. You can enable
debugging for your build for the subsystem or feature by using the ``menuconfig`` system.
The debug options are available under :menuselection:`Build Setup --> Debug Options`. You will most likely have to enable the
following options:
* :menuselection:`Enable Debug Features` — selecting this will turn on subsystem-level debugging options, they will become visible
on the page below. You can then select the ones you want.
* :menuselection:`Enable Error Output` — this will only log errors.
* :menuselection:`Enable Warnings Output` — this will log warnings and errors.
* :menuselection:`Enable Informational Debug Output` — this will produce informational output, warnings, and errors.
You can then select from the subsystems that are available, Network, Scheduler, USB, etc. Note that you will need to
separately enable the subsystem elsewhere in the ``menuconfig`` system. To see the ``CONFIG`` define that is set,
use the arrow keys to highlight the subsystem (for instance, :menuselection:`Network Debug Features`) and type :kbd:`?`. This will show
you that the C macro that is set is called ``CONFIG_DEBUG_NET``. ``debug.h`` defines the ``netinfo()`` logging
function that will log output if this macro is set. You can search the source code for ``netinfo`` to see how it is
used.
.. image:: ../_static/images/menuconfig-debug.png
:width: 800px
:align: center
:alt: Screenshot of menuconfig system main screen
Note that enabling all these will produce an incredible amount of logging output. Enable the level you want and
the area you're interested in, and leave the rest disabled, save the config, and then recompile. You can see the full
list of debug feature logging functions in the file
`debug.h <https://github.com/apache/nuttx/blob/master/include/debug.h>`__.
Syslog timestamps can be enabled in the configuration in :menuselection:`Device Drivers --> System Logging --> Prepend
timestamp to syslog message` (``CONFIG_SYSLOG_TIMESTAMP``).
You may need to do a little bit of experimenting to find the combination of logging settings that work for the problem
you're trying to solve. See the file `debug.h <https://github.com/apache/nuttx/blob/master/include/debug.h>`_
for available debug settings that are available.
There are also subsystems that enable USB trace debugging, and you can log to memory too, if you need the logging to be
faster than what the console can output.
Debugging with ``openocd`` and ``gdb``
======================================
To debug our Nucleo board using its embedded SWD debug adapter,
start ``openocd`` with the following command:
.. code-block:: console
$ openocd -f interface/stlink-v2.cfg -f target/stm32f1x.cfg
This will start a ``gdb`` server. Then, start ``gdb`` with:
.. code-block:: console
$ cd nuttx/
$ gdb-multiarch nuttx/nuttx
Inside ``gdb`` console, connect to the ``gdb`` server with:
.. code-block::
(gdb) target extended-remote :3333
You can now use standard ``gdb`` commands. For example, to
reset the board:
.. code-block::
(gdb) mon reset
To halt the board:
.. code-block::
(gdb) mon halt
To set a breakpoint:
.. code-block::
(gdb) breakpoint nsh_main
and to finally start nuttx:
.. code-block::
(gdb) continue
Continuing.
Breakpoint 1, nsh_main (argc=1, argv=0x200ddfac) at nsh_main.c:208
208 sched_getparam(0, &param);
(gdb) continue
Continuing.
.. tip::
You can abbreviate ``gdb`` commands: ``info b`` is a shortcut for
``information breakpoints``; ``c`` works the same as ``continue``, etc.
NuttX aware debugging
---------------------
Since NuttX is actually an RTOS, it is useful to have ``gdb`` be aware of the different
tasks/threads that are running. There are two ways to do this: via ``openocd``
itself or via ``gdb``. Note that in both cases, you need to enable debug symbols
(``CONFIG_DEBUG_SYMBOLS``).
With openocd
~~~~~~~~~~~~
``openocd`` supports various RTOS directly, including NuttX. It works by reading
into internal NuttX symbols which define the active tasks and their properties.
As a result, the ``gdb`` server will directly be aware of each task as a different
`thread`. The downside of this approach is that it depends on how you build NuttX
as there are some options hardcoded into
opencd. By default, it assumes:
* ``CONFIG_DISABLE_MQUEUE=y``
* ``CONFIG_PAGING=n``
If you need these options to be set differently, you will have to edit ``./src/rtos/nuttx_header.h`` from ``openocd``,
change the corresponding settings and then rebuild it.
Finally, to enable NuttX integration, you need to supply an additional ``openocd`` argument:
.. code-block:: console
$ openocd -f interface/stlink-v2.cfg -f target/stm32f1x.cfg -c '$_TARGETNAME configure -rtos nuttx'
Since ``openocd`` also needs to know the memory layout of certain datastructures, you need to have ``gdb``
run the following commands once the ``nuttx`` binary is loaded:
.. code-block::
eval "monitor nuttx.pid_offset %d", &((struct tcb_s *)(0))->pid
eval "monitor nuttx.xcpreg_offset %d", &((struct tcb_s *)(0))->xcp.regs
eval "monitor nuttx.state_offset %d", &((struct tcb_s *)(0))->task_state
eval "monitor nuttx.name_offset %d", &((struct tcb_s *)(0))->name
eval "monitor nuttx.name_size %d", sizeof(((struct tcb_s *)(0))->name)
One way to do this is to define a gdb `hook` function that will be called when running ``file`` command:
.. code-block::
define hookpost-file
eval "monitor nuttx.pid_offset %d", &((struct tcb_s *)(0))->pid
eval "monitor nuttx.xcpreg_offset %d", &((struct tcb_s *)(0))->xcp.regs
eval "monitor nuttx.state_offset %d", &((struct tcb_s *)(0))->task_state
eval "monitor nuttx.name_offset %d", &((struct tcb_s *)(0))->name
eval "monitor nuttx.name_size %d", sizeof(((struct tcb_s *)(0))->name)
end
You will see that ``openocd`` has received the memory offsets in its output:
.. code-block::
Open On-Chip Debugger 0.10.0+dev-01514-ga8edbd020-dirty (2020-11-20-14:23)
Licensed under GNU GPL v2
For bug reports, read
http://openocd.org/doc/doxygen/bugs.html
Info : auto-selecting first available session transport "swd". To override use 'transport select <transport>'.
Info : target type name = cortex_m
Info : Listening on port 6666 for tcl connections
Info : Listening on port 4444 for telnet connections
15:41:23: Debugging starts
Info : CMSIS-DAP: SWD Supported
Info : CMSIS-DAP: FW Version = 1.10
Info : CMSIS-DAP: Interface Initialised (SWD)
Info : SWCLK/TCK = 1 SWDIO/TMS = 1 TDI = 0 TDO = 0 nTRST = 0 nRESET = 1
Info : CMSIS-DAP: Interface ready
Info : clock speed 1000 kHz
Info : SWD DPIDR 0x2ba01477
Info : nrf52.cpu: hardware has 6 breakpoints, 4 watchpoints
Info : starting gdb server for nrf52.cpu on 3333
Info : Listening on port 3333 for gdb connections
Info : accepting 'gdb' connection on tcp/3333
Error: No symbols for NuttX
Info : nRF52832-QFAA(build code: B0) 512kB Flash, 64kB RAM
undefined debug reason 8 - target needs reset
Warn : Prefer GDB command "target extended-remote 3333" instead of "target remote 3333"
Info : pid_offset: 12
Info : xcpreg_offset: 132
Info : state_offset: 26
Info : name_offset: 208
Info : name_size: 32
target halted due to debug-request, current mode: Thread
xPSR: 0x01000000 pc: 0x000000dc msp: 0x20000cf0
target halted due to debug-request, current mode: Thread xPSR: 0x01000000 pc: 0x000000dc msp: 0x20000cf0
.. note:: You will probably see the ``Error: No symbols for NuttX`` error appear once at startup. This is OK
unless you see it every time you step the debugger. In this case, it would mean you did not enable debug symbols.
Now, You can now inspect threads:
.. code-block::
(gdb) info threads
Id Target Id Frame
* 1 Remote target nx_start_application () at init/nx_bringup.c:261
(gdb) info registers
r0 0x0 0
r1 0x2f 47
r2 0x0 0
r3 0x0 0
r4 0x0 0
r5 0x0 0
r6 0x0 0
r7 0x20000ca0 536874144
r8 0x0 0
r9 0x0 0
r10 0x0 0
r11 0x0 0
r12 0x9 9
sp 0x20000c98 0x20000c98
lr 0x19c5 6597
pc 0x1996 0x1996 <nx_start_application+10>
xPSR 0x41000000 1090519040
fpscr 0x0 0
msp 0x20000c98 0x20000c98
psp 0x0 0x0 <_vectors>
primask 0x0 0
basepri 0xe0 -32
faultmask 0x0 0
control 0x0 0
With gdb
~~~~~~~~
You can also do NuttX aware debugging using ``gdb`` scripting support.
The benefit is that it works also for the sim build where ``openocd`` is
not applicable. For this to work, you will need to enable PROC filesystem support
which will expose required task information (``CONFIG_FS_PROCFS=y CONFIG_DEBUG_TCBINFO=y``).
To use this approach, you can load the ``nuttx/tools/gdb/__init__.py`` file. An
easy way to do this is to add an extra command:
.. code-block:: console
$ gdb nuttx -ex "nuttx/tools/gdb/__init__.py"
gdb can need to set the current elf support architecture, for example,
the prefix is arm-ebai-none-.
.. code-block::
(gdb) info threads
Id Thread Info Frame
*0 Thread 0x20000398 (Name: Idle Task, State: Running, Priority: 0, Stack: 1000) 0x80001ac __start() at chip/stm32_start.c:111
1 Thread 0x10000188 (Name: nsh_main, State: Waiting,Semaphore, Priority: 100, Stack: 2000) 0x800aa06 sys_call2() at /home/ajh/work/vela_all/nuttx/include/arch/syscall.h:187
The python script has extended many commands like ``thread <id>`` ,
``thread apply <all|id list> cmd``, ``nxsetargs`` etc.
You can use ``help <command>`` to get help.
Note that if you need to continue debugging after using the thread command,
please use ``c`` instead of ``continue``, because thread will force the register to be set,
and the `c` command will restore the register before conitune.