The Slick Debugger
Ensure you have the following dependencies:
- Python 3.6 or newer
- libkdumpfile (optional - needed for kdump-compressed crash dumps)
- drgn
Note that in order for drgn
to support kdump files it needs to be compiled with libkdumpfile
. Unfortunately that means that users should always install libkdumpfile
first before installing drgn
.
Finally run the following to install sdb
:
$ git clone https://github.com/delphix/sdb.git
$ cd sdb
$ sudo python3 setup.py install
The above should install sdb
under /usr/local/bin/
.
Running sudo sdb
attaches sdb to the running kernel by default.
To debug a running program, run sudo sdb -p <PID>
.
For post-mortem debugging (either a kernel crash dump or a userland core dump), use sudo sdb <vmlinux path|userland binary path> <dump>
.
$ sudo sdb
sdb> find_task 1 | member comm
(char [16])"systemd"
sdb> find_task 1 | stack
TASK_STRUCT STATE COUNT
==========================================
0xffff89cea441dd00 INTERRUPTIBLE 1
__schedule+0x2e5
schedule+0x33
schedule_hrtimeout_range_clock+0xfd
schedule_hrtimeout_range+0x13
ep_poll+0x40a
do_epoll_wait+0xb7
__x64_sys_epoll_wait+0x1e
do_syscall_64+0x57
entry_SYSCALL_64+0x7c
sdb> addr modules | lxlist "struct module" list | member name ! sort | head -n 3
(char [56])"aesni_intel"
(char [56])"async_memcpy"
(char [56])"async_pq"
$ python3 -m pip install pylint pytest
$ python3 -m pylint -d duplicate-code -d invalid-name sdb
$ python3 -m pylint -d duplicate-code -d invalid-name tests
$ python3 -m pip install mypy==0.730
$ python3 -m mypy --strict --show-error-codes -p sdb
$ python3 -m mypy --strict --ignore-missing-imports --show-error-codes -p tests
$ python3 -m pip install yapf
$ python3 -m yapf --diff --style google --recursive sdb
$ python3 -m yapf --diff --style google --recursive tests
If yapf
has suggestions you can apply them automatically by substituting
--diff
with -i
like this:
$ python3 -m yapf -i --style google --recursive sdb
$ python3 -m yapf -i --style google --recursive tests
Regression testing is currently done by downloading a refererence crash/core
dump and running a predetermined set of commads on it, ensuring that they
return the same reference output before and after your changes. To see the list
of reference crash dumps refer to the testing matrix of the pytest
Github
action in .github/workflows/main.yml
. Here is an example of running the
regression test commands against dump.201912060006.tar.lzma
with code
coverage and verbose output:
$ python3 -m pip install python-config pytest pytest-cov
$ .github/scripts/download-dump-from-s3.sh dump.201912060006.tar.lzma
$ python3 -m pytest -v --cov sdb --cov-report xml tests
If you want pytest
to stop on the first failure it encounters add
-x/--exitfirst
in the command above.
If you've added new test commands or found mistakes in the current reference output and you want (re)generate some reference output download all crash/core dumps (or the specific one you want to correct) and run the following:
$ PYTHONPATH=$(pwd) python3 tests/integration/gen_regression_output.py