1 .. SPDX-License-Identifier: GPL-2.0
3 ==========================
4 Userspace debugging advice
5 ==========================
7 This document provides a brief overview of common tools to debug the Linux
9 For debugging advice aimed at driver developers go :doc:`here
10 </process/debugging/driver_development_debugging_guide>`.
11 For general debugging advice, see :doc:`general advice document
12 </process/debugging/index>`.
17 The following sections show you the available tools.
22 Mechanism to filter what ends up in the kernel log by dis-/en-abling log
25 Prerequisite: ``CONFIG_DYNAMIC_DEBUG``
27 Dynamic debug is only able to target:
31 - print_hex_dump_debug()
32 - print_hex_dump_bytes()
34 Therefore the usability of this tool is, as of now, quite limited as there is
35 no uniform rule for adding debug prints to the codebase, resulting in a variety
36 of ways these prints are implemented.
38 Also, note that most debug statements are implemented as a variation of
39 dprintk(), which have to be activated via a parameter in respective module,
40 dynamic debug is unable to do that step for you.
42 Here is one example, that enables all available pr_debug()'s within the file::
44 $ alias ddcmd='echo $* > /proc/dynamic_debug/control'
45 $ ddcmd '-p; file v4l2-h264.c +p'
46 $ grep =p /proc/dynamic_debug/control
47 drivers/media/v4l2-core/v4l2-h264.c:372 [v4l2_h264]print_ref_list_b =p
48 "ref_pic_list_b%u (cur_poc %u%c) %s"
49 drivers/media/v4l2-core/v4l2-h264.c:333 [v4l2_h264]print_ref_list_p =p
50 "ref_pic_list_p (cur_poc %u%c) %s\n"
52 **When should you use this over Ftrace ?**
54 - When the code contains one of the valid print statements (see above) or when
55 you have added multiple pr_debug() statements during development
56 - When timing is not an issue, meaning if multiple pr_debug() statements in
57 the code won't cause delays
58 - When you care more about receiving specific log messages than tracing the
59 pattern of how a function is called
61 For the full documentation see :doc:`/admin-guide/dynamic-debug-howto`
66 Prerequisite: ``CONFIG_DYNAMIC_FTRACE``
68 This tool uses the tracefs file system for the control files and output files.
69 That file system will be mounted as a ``tracing`` directory, which can be found
70 in either ``/sys/kernel/`` or ``/sys/debug/kernel/``.
72 Some of the most important operations for debugging are:
74 - You can perform a function trace by adding a function name to the
75 ``set_ftrace_filter`` file (which accepts any function name found within the
76 ``available_filter_functions`` file) or you can specifically disable certain
77 functions by adding their names to the ``set_ftrace_notrace`` file (more info
78 at: :ref:`trace/ftrace:dynamic ftrace`).
79 - In order to find out where calls originate from you can activate the
80 ``func_stack_trace`` option under ``options/func_stack_trace``.
81 - Tracing the children of a function call and showing the return values are
82 possible by adding the desired function in the ``set_graph_function`` file
83 (requires config ``FUNCTION_GRAPH_RETVAL``); more info at
84 :ref:`trace/ftrace:dynamic ftrace with the function graph tracer`.
86 For the full Ftrace documentation see :doc:`/trace/ftrace`
88 Or you could also trace for specific events by :ref:`using event tracing
89 <trace/events:2. using event tracing>`, which can be defined as described here:
90 :ref:`Creating a custom Ftrace tracepoint
91 <process/debugging/driver_development_debugging_guide:ftrace>`.
93 For the full Ftrace event tracing documentation see :doc:`/trace/events`
97 Reading the ftrace log
98 ~~~~~~~~~~~~~~~~~~~~~~
100 The ``trace`` file can be read just like any other file (``cat``, ``tail``,
101 ``head``, ``vim``, etc.), the size of the file is limited by the
102 ``buffer_size_kb`` (``echo 1000 > buffer_size_kb``). The
103 :ref:`trace/ftrace:trace_pipe` will behave similarly to the ``trace`` file, but
104 whenever you read from the file the content is consumed.
109 A GUI interface to visualize the traces as a graph and list view from the
110 output of the `trace-cmd
111 <https://git.kernel.org/pub/scm/utils/trace-cmd/trace-cmd.git/>`__ application.
113 For the full documentation see `<https://kernelshark.org/Documentation.html>`__
118 The tools mentioned above provide ways to inspect kernel code, results,
119 variable values, etc. Sometimes you have to find out first where to look and
120 for those cases, a box of performance tracking tools can help you to frame the
123 Why should you do a performance analysis?
124 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
126 A performance analysis is a good first step when among other reasons:
128 - you cannot define the issue
129 - you do not know where it occurs
130 - the running system should not be interrupted or it is a remote system, where
131 you cannot install a new module/kernel
133 How to do a simple analysis with linux tools?
134 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
136 For the start of a performance analysis, you can start with the usual tools
139 - ``top`` / ``htop`` / ``atop`` (*get an overview of the system load, see
140 spikes on specific processes*)
141 - ``mpstat -P ALL`` (*look at the load distribution among CPUs*)
142 - ``iostat -x`` (*observe input and output devices utilization and performance*)
143 - ``vmstat`` (*overview of memory usage on the system*)
144 - ``pidstat`` (*similar to* ``vmstat`` *but per process, to dial it down to the
146 - ``strace -tp $PID`` (*once you know the process, you can figure out how it
147 communicates with the Kernel*)
149 These should help to narrow down the areas to look at sufficiently.
151 Diving deeper with perf
152 ~~~~~~~~~~~~~~~~~~~~~~~
154 The **perf** tool provides a series of metrics and events to further dial down
157 Prerequisite: build or install perf on your system
159 Gather statistics data for finding all files starting with ``gcc`` in ``/usr``::
161 # perf stat -d find /usr -name 'gcc*' | wc -l
163 Performance counter stats for 'find /usr -name gcc*':
165 1277.81 msec task-clock # 0.997 CPUs utilized
166 9 context-switches # 7.043 /sec
167 1 cpu-migrations # 0.783 /sec
168 704 page-faults # 550.943 /sec
169 766548897 cycles # 0.600 GHz (97.15%)
170 798285467 instructions # 1.04 insn per cycle (97.15%)
171 57582731 branches # 45.064 M/sec (2.85%)
172 3842573 branch-misses # 6.67% of all branches (97.15%)
173 281616097 L1-dcache-loads # 220.390 M/sec (97.15%)
174 4220975 L1-dcache-load-misses # 1.50% of all L1-dcache accesses (97.15%)
175 <not supported> LLC-loads
176 <not supported> LLC-load-misses
178 1.281746009 seconds time elapsed
180 0.508796000 seconds user
181 0.773209000 seconds sys
186 The availability of events and metrics depends on the system you are running.
188 For the full documentation see
189 `<https://perf.wiki.kernel.org/index.php/Main_Page>`__
194 A set of tools to measure and analyze how well applications and systems perform.
197 * identify bottlenecks
199 * make software run faster and more efficiently.
201 **What is the difference between perfetto and perf?**
203 * perf is tool as part of and specialized for the Linux Kernel and has CLI user
205 * perfetto cross-platform performance analysis stack, has extended
206 functionality into userspace and provides a WEB user interface.
208 For the full documentation see `<https://perfetto.dev/docs/>`__
210 Kernel panic analysis tools
211 ---------------------------
213 To capture the crash dump please use ``Kdump`` & ``Kexec``. Below you can find
214 some advice for analysing the data.
216 For the full documentation see the :doc:`/admin-guide/kdump/kdump`
218 In order to find the corresponding line in the code you can use `faddr2line
219 <https://elixir.bootlin.com/linux/v6.11.6/source/scripts/faddr2line>`__; note
220 that you need to enable ``CONFIG_DEBUG_INFO`` for that to work.
222 An alternative to using ``faddr2line`` is the use of ``objdump`` (and its
223 derivatives for the different platforms like ``aarch64-linux-gnu-objdump``).
224 Take this line as an example:
226 ``[ +0.000240] rkvdec_device_run+0x50/0x138 [rockchip_vdec]``.
228 We can find the corresponding line of code by executing::
230 aarch64-linux-gnu-objdump -dS drivers/staging/media/rkvdec/rockchip-vdec.ko | grep rkvdec_device_run\>: -A 40
231 0000000000000ac8 <rkvdec_device_run>:
235 ad0: d503233f paciasp
236 ad4: a9bd7bfd stp x29, x30, [sp, #-48]!
237 ad8: 910003fd mov x29, sp
238 adc: a90153f3 stp x19, x20, [sp, #16]
239 ae0: a9025bf5 stp x21, x22, [sp, #32]
240 const struct rkvdec_coded_fmt_desc *desc = ctx->coded_fmt_desc;
241 ae4: f9411814 ldr x20, [x0, #560]
242 struct rkvdec_dev *rkvdec = ctx->dev;
243 ae8: f9418015 ldr x21, [x0, #768]
245 aec: b4000654 cbz x20, bb4 <rkvdec_device_run+0xec>
246 ret = pm_runtime_resume_and_get(rkvdec->dev);
247 af0: f943d2b6 ldr x22, [x21, #1952]
248 ret = __pm_runtime_resume(dev, RPM_GET_PUT);
249 af4: aa0003f3 mov x19, x0
250 af8: 52800081 mov w1, #0x4 // #4
251 afc: aa1603e0 mov x0, x22
252 b00: 94000000 bl 0 <__pm_runtime_resume>
254 b04: 37f80340 tbnz w0, #31, b6c <rkvdec_device_run+0xa4>
255 dev_warn(rkvdec->dev, "Not good\n");
256 b08: f943d2a0 ldr x0, [x21, #1952]
257 b0c: 90000001 adrp x1, 0 <rkvdec_try_ctrl-0x8>
258 b10: 91000021 add x1, x1, #0x0
259 b14: 94000000 bl 0 <_dev_warn>
261 b18: d2800001 mov x1, #0x0 // #0
264 Meaning, in this line from the crash dump::
266 [ +0.000240] rkvdec_device_run+0x50/0x138 [rockchip_vdec]
268 I can take the ``0x50`` as offset, which I have to add to the base address
269 of the corresponding function, which I find in this line::
271 0000000000000ac8 <rkvdec_device_run>:
273 The result of ``0xac8 + 0x50 = 0xb18``
274 And when I search for that address within the function I get the
278 b18: d2800001 mov x1, #0x0
280 **Copyright** ©2024 : Collabora