xref: /linux/tools/bpf/bpftool/Documentation/bpftool-prog.rst (revision 4b660dbd9ee2059850fd30e0df420ca7a38a1856)
1.. SPDX-License-Identifier: (GPL-2.0-only OR BSD-2-Clause)
2
3================
4bpftool-prog
5================
6-------------------------------------------------------------------------------
7tool for inspection and simple manipulation of eBPF progs
8-------------------------------------------------------------------------------
9
10:Manual section: 8
11
12.. include:: substitutions.rst
13
14SYNOPSIS
15========
16
17	**bpftool** [*OPTIONS*] **prog** *COMMAND*
18
19	*OPTIONS* := { |COMMON_OPTIONS| |
20	{ **-f** | **--bpffs** } | { **-m** | **--mapcompat** } | { **-n** | **--nomount** } |
21	{ **-L** | **--use-loader** } }
22
23	*COMMANDS* :=
24	{ **show** | **list** | **dump xlated** | **dump jited** | **pin** | **load** |
25	**loadall** | **help** }
26
27PROG COMMANDS
28=============
29
30|	**bpftool** **prog** { **show** | **list** } [*PROG*]
31|	**bpftool** **prog dump xlated** *PROG* [{ **file** *FILE* | [**opcodes**] [**linum**] [**visual**] }]
32|	**bpftool** **prog dump jited**  *PROG* [{ **file** *FILE* | [**opcodes**] [**linum**] }]
33|	**bpftool** **prog pin** *PROG* *FILE*
34|	**bpftool** **prog** { **load** | **loadall** } *OBJ* *PATH* [**type** *TYPE*] [**map** { **idx** *IDX* | **name** *NAME* } *MAP*] [{ **offload_dev** | **xdpmeta_dev** } *NAME*] [**pinmaps** *MAP_DIR*] [**autoattach**]
35|	**bpftool** **prog attach** *PROG* *ATTACH_TYPE* [*MAP*]
36|	**bpftool** **prog detach** *PROG* *ATTACH_TYPE* [*MAP*]
37|	**bpftool** **prog tracelog**
38|	**bpftool** **prog run** *PROG* **data_in** *FILE* [**data_out** *FILE* [**data_size_out** *L*]] [**ctx_in** *FILE* [**ctx_out** *FILE* [**ctx_size_out** *M*]]] [**repeat** *N*]
39|	**bpftool** **prog profile** *PROG* [**duration** *DURATION*] *METRICs*
40|	**bpftool** **prog help**
41|
42|	*MAP* := { **id** *MAP_ID* | **pinned** *FILE* }
43|	*PROG* := { **id** *PROG_ID* | **pinned** *FILE* | **tag** *PROG_TAG* | **name** *PROG_NAME* }
44|	*TYPE* := {
45|		**socket** | **kprobe** | **kretprobe** | **classifier** | **action** |
46|		**tracepoint** | **raw_tracepoint** | **xdp** | **perf_event** | **cgroup/skb** |
47|		**cgroup/sock** | **cgroup/dev** | **lwt_in** | **lwt_out** | **lwt_xmit** |
48|		**lwt_seg6local** | **sockops** | **sk_skb** | **sk_msg** | **lirc_mode2** |
49|		**cgroup/bind4** | **cgroup/bind6** | **cgroup/post_bind4** | **cgroup/post_bind6** |
50|		**cgroup/connect4** | **cgroup/connect6** | **cgroup/connect_unix** |
51|		**cgroup/getpeername4** | **cgroup/getpeername6** | **cgroup/getpeername_unix** |
52|		**cgroup/getsockname4** | **cgroup/getsockname6** | **cgroup/getsockname_unix** |
53|		**cgroup/sendmsg4** | **cgroup/sendmsg6** | **cgroup/sendmsg_unix** |
54|		**cgroup/recvmsg4** | **cgroup/recvmsg6** | **cgroup/recvmsg_unix** | **cgroup/sysctl** |
55|		**cgroup/getsockopt** | **cgroup/setsockopt** | **cgroup/sock_release** |
56|		**struct_ops** | **fentry** | **fexit** | **freplace** | **sk_lookup**
57|	}
58|	*ATTACH_TYPE* := {
59|		**sk_msg_verdict** | **sk_skb_verdict** | **sk_skb_stream_verdict** |
60|		**sk_skb_stream_parser** | **flow_dissector**
61|	}
62|	*METRICs* := {
63|		**cycles** | **instructions** | **l1d_loads** | **llc_misses** |
64|		**itlb_misses** | **dtlb_misses**
65|	}
66
67
68DESCRIPTION
69===========
70	**bpftool prog { show | list }** [*PROG*]
71		  Show information about loaded programs.  If *PROG* is
72		  specified show information only about given programs,
73		  otherwise list all programs currently loaded on the system.
74		  In case of **tag** or **name**, *PROG* may match several
75		  programs which will all be shown.
76
77		  Output will start with program ID followed by program type and
78		  zero or more named attributes (depending on kernel version).
79
80		  Since Linux 5.1 the kernel can collect statistics on BPF
81		  programs (such as the total time spent running the program,
82		  and the number of times it was run). If available, bpftool
83		  shows such statistics. However, the kernel does not collect
84		  them by defaults, as it slightly impacts performance on each
85		  program run. Activation or deactivation of the feature is
86		  performed via the **kernel.bpf_stats_enabled** sysctl knob.
87
88		  Since Linux 5.8 bpftool is able to discover information about
89		  processes that hold open file descriptors (FDs) against BPF
90		  programs. On such kernels bpftool will automatically emit this
91		  information as well.
92
93	**bpftool prog dump xlated** *PROG* [{ **file** *FILE* | [**opcodes**] [**linum**] [**visual**] }]
94		  Dump eBPF instructions of the programs from the kernel. By
95		  default, eBPF will be disassembled and printed to standard
96		  output in human-readable format. In this case, **opcodes**
97		  controls if raw opcodes should be printed as well.
98
99		  In case of **tag** or **name**, *PROG* may match several
100		  programs which will all be dumped.  However, if **file** or
101		  **visual** is specified, *PROG* must match a single program.
102
103		  If **file** is specified, the binary image will instead be
104		  written to *FILE*.
105
106		  If **visual** is specified, control flow graph (CFG) will be
107		  built instead, and eBPF instructions will be presented with
108		  CFG in DOT format, on standard output.
109
110		  If the programs have line_info available, the source line will
111		  be displayed.  If **linum** is specified, the filename, line
112		  number and line column will also be displayed.
113
114	**bpftool prog dump jited**  *PROG* [{ **file** *FILE* | [**opcodes**] [**linum**] }]
115		  Dump jited image (host machine code) of the program.
116
117		  If *FILE* is specified image will be written to a file,
118		  otherwise it will be disassembled and printed to stdout.
119		  *PROG* must match a single program when **file** is specified.
120
121		  **opcodes** controls if raw opcodes will be printed.
122
123		  If the prog has line_info available, the source line will
124		  be displayed.  If **linum** is specified, the filename, line
125		  number and line column will also be displayed.
126
127	**bpftool prog pin** *PROG* *FILE*
128		  Pin program *PROG* as *FILE*.
129
130		  Note: *FILE* must be located in *bpffs* mount. It must not
131		  contain a dot character ('.'), which is reserved for future
132		  extensions of *bpffs*.
133
134	**bpftool prog { load | loadall }** *OBJ* *PATH* [**type** *TYPE*] [**map** { **idx** *IDX* | **name** *NAME* } *MAP*] [{ **offload_dev** | **xdpmeta_dev** } *NAME*] [**pinmaps** *MAP_DIR*] [**autoattach**]
135		  Load bpf program(s) from binary *OBJ* and pin as *PATH*.
136		  **bpftool prog load** pins only the first program from the
137		  *OBJ* as *PATH*. **bpftool prog loadall** pins all programs
138		  from the *OBJ* under *PATH* directory.
139		  **type** is optional, if not specified program type will be
140		  inferred from section names.
141		  By default bpftool will create new maps as declared in the ELF
142		  object being loaded.  **map** parameter allows for the reuse
143		  of existing maps.  It can be specified multiple times, each
144		  time for a different map.  *IDX* refers to index of the map
145		  to be replaced in the ELF file counting from 0, while *NAME*
146		  allows to replace a map by name.  *MAP* specifies the map to
147		  use, referring to it by **id** or through a **pinned** file.
148		  If **offload_dev** *NAME* is specified program will be loaded
149		  onto given networking device (offload).
150		  If **xdpmeta_dev** *NAME* is specified program will become
151		  device-bound without offloading, this facilitates access
152		  to XDP metadata.
153		  Optional **pinmaps** argument can be provided to pin all
154		  maps under *MAP_DIR* directory.
155
156		  If **autoattach** is specified program will be attached
157		  before pin. In that case, only the link (representing the
158		  program attached to its hook) is pinned, not the program as
159		  such, so the path won't show in **bpftool prog show -f**,
160		  only show in **bpftool link show -f**. Also, this only works
161		  when bpftool (libbpf) is able to infer all necessary
162		  information from the object file, in particular, it's not
163		  supported for all program types. If a program does not
164		  support autoattach, bpftool falls back to regular pinning
165		  for that program instead.
166
167		  Note: *PATH* must be located in *bpffs* mount. It must not
168		  contain a dot character ('.'), which is reserved for future
169		  extensions of *bpffs*.
170
171	**bpftool prog attach** *PROG* *ATTACH_TYPE* [*MAP*]
172		  Attach bpf program *PROG* (with type specified by
173		  *ATTACH_TYPE*). Most *ATTACH_TYPEs* require a *MAP*
174		  parameter, with the exception of *flow_dissector* which is
175		  attached to current networking name space.
176
177	**bpftool prog detach** *PROG* *ATTACH_TYPE* [*MAP*]
178		  Detach bpf program *PROG* (with type specified by
179		  *ATTACH_TYPE*). Most *ATTACH_TYPEs* require a *MAP*
180		  parameter, with the exception of *flow_dissector* which is
181		  detached from the current networking name space.
182
183	**bpftool prog tracelog**
184		  Dump the trace pipe of the system to the console (stdout).
185		  Hit <Ctrl+C> to stop printing. BPF programs can write to this
186		  trace pipe at runtime with the **bpf_trace_printk**\ () helper.
187		  This should be used only for debugging purposes. For
188		  streaming data from BPF programs to user space, one can use
189		  perf events (see also **bpftool-map**\ (8)).
190
191	**bpftool prog run** *PROG* **data_in** *FILE* [**data_out** *FILE* [**data_size_out** *L*]] [**ctx_in** *FILE* [**ctx_out** *FILE* [**ctx_size_out** *M*]]] [**repeat** *N*]
192		  Run BPF program *PROG* in the kernel testing infrastructure
193		  for BPF, meaning that the program works on the data and
194		  context provided by the user, and not on actual packets or
195		  monitored functions etc. Return value and duration for the
196		  test run are printed out to the console.
197
198		  Input data is read from the *FILE* passed with **data_in**.
199		  If this *FILE* is "**-**", input data is read from standard
200		  input. Input context, if any, is read from *FILE* passed with
201		  **ctx_in**. Again, "**-**" can be used to read from standard
202		  input, but only if standard input is not already in use for
203		  input data. If a *FILE* is passed with **data_out**, output
204		  data is written to that file. Similarly, output context is
205		  written to the *FILE* passed with **ctx_out**. For both
206		  output flows, "**-**" can be used to print to the standard
207		  output (as plain text, or JSON if relevant option was
208		  passed). If output keywords are omitted, output data and
209		  context are discarded. Keywords **data_size_out** and
210		  **ctx_size_out** are used to pass the size (in bytes) for the
211		  output buffers to the kernel, although the default of 32 kB
212		  should be more than enough for most cases.
213
214		  Keyword **repeat** is used to indicate the number of
215		  consecutive runs to perform. Note that output data and
216		  context printed to files correspond to the last of those
217		  runs. The duration printed out at the end of the runs is an
218		  average over all runs performed by the command.
219
220		  Not all program types support test run. Among those which do,
221		  not all of them can take the **ctx_in**/**ctx_out**
222		  arguments. bpftool does not perform checks on program types.
223
224	**bpftool prog profile** *PROG* [**duration** *DURATION*] *METRICs*
225		  Profile *METRICs* for bpf program *PROG* for *DURATION*
226		  seconds or until user hits <Ctrl+C>. *DURATION* is optional.
227		  If *DURATION* is not specified, the profiling will run up to
228		  **UINT_MAX** seconds.
229
230	**bpftool prog help**
231		  Print short help message.
232
233OPTIONS
234=======
235	.. include:: common_options.rst
236
237	-f, --bpffs
238		  When showing BPF programs, show file names of pinned
239		  programs.
240
241	-m, --mapcompat
242		  Allow loading maps with unknown map definitions.
243
244	-n, --nomount
245		  Do not automatically attempt to mount any virtual file system
246		  (such as tracefs or BPF virtual file system) when necessary.
247
248	-L, --use-loader
249		  Load program as a "loader" program. This is useful to debug
250		  the generation of such programs. When this option is in
251		  use, bpftool attempts to load the programs from the object
252		  file into the kernel, but does not pin them (therefore, the
253		  *PATH* must not be provided).
254
255		  When combined with the **-d**\ \|\ **--debug** option,
256		  additional debug messages are generated, and the execution
257		  of the loader program will use the **bpf_trace_printk**\ ()
258		  helper to log each step of loading BTF, creating the maps,
259		  and loading the programs (see **bpftool prog tracelog** as
260		  a way to dump those messages).
261
262EXAMPLES
263========
264**# bpftool prog show**
265
266::
267
268    10: xdp  name some_prog  tag 005a3d2123620c8b  gpl run_time_ns 81632 run_cnt 10
269            loaded_at 2017-09-29T20:11:00+0000  uid 0
270            xlated 528B  jited 370B  memlock 4096B  map_ids 10
271            pids systemd(1)
272
273**# bpftool --json --pretty prog show**
274
275::
276
277    [{
278            "id": 10,
279            "type": "xdp",
280            "tag": "005a3d2123620c8b",
281            "gpl_compatible": true,
282            "run_time_ns": 81632,
283            "run_cnt": 10,
284            "loaded_at": 1506715860,
285            "uid": 0,
286            "bytes_xlated": 528,
287            "jited": true,
288            "bytes_jited": 370,
289            "bytes_memlock": 4096,
290            "map_ids": [10
291            ],
292            "pids": [{
293                    "pid": 1,
294                    "comm": "systemd"
295                }
296            ]
297        }
298    ]
299
300|
301| **# bpftool prog dump xlated id 10 file /tmp/t**
302| **$ ls -l /tmp/t**
303
304::
305
306    -rw------- 1 root root 560 Jul 22 01:42 /tmp/t
307
308**# bpftool prog dump jited tag 005a3d2123620c8b**
309
310::
311
312    0:   push   %rbp
313    1:   mov    %rsp,%rbp
314    2:   sub    $0x228,%rsp
315    3:   sub    $0x28,%rbp
316    4:   mov    %rbx,0x0(%rbp)
317
318|
319| **# mount -t bpf none /sys/fs/bpf/**
320| **# bpftool prog pin id 10 /sys/fs/bpf/prog**
321| **# bpftool prog load ./my_prog.o /sys/fs/bpf/prog2**
322| **# ls -l /sys/fs/bpf/**
323
324::
325
326    -rw------- 1 root root 0 Jul 22 01:43 prog
327    -rw------- 1 root root 0 Jul 22 01:44 prog2
328
329**# bpftool prog dump jited pinned /sys/fs/bpf/prog opcodes**
330
331::
332
333   0:   push   %rbp
334        55
335   1:   mov    %rsp,%rbp
336        48 89 e5
337   4:   sub    $0x228,%rsp
338        48 81 ec 28 02 00 00
339   b:   sub    $0x28,%rbp
340        48 83 ed 28
341   f:   mov    %rbx,0x0(%rbp)
342        48 89 5d 00
343
344|
345| **# bpftool prog load xdp1_kern.o /sys/fs/bpf/xdp1 type xdp map name rxcnt id 7**
346| **# bpftool prog show pinned /sys/fs/bpf/xdp1**
347
348::
349
350    9: xdp  name xdp_prog1  tag 539ec6ce11b52f98  gpl
351            loaded_at 2018-06-25T16:17:31-0700  uid 0
352            xlated 488B  jited 336B  memlock 4096B  map_ids 7
353
354**# rm /sys/fs/bpf/xdp1**
355
356|
357| **# bpftool prog profile id 337 duration 10 cycles instructions llc_misses**
358
359::
360
361         51397 run_cnt
362      40176203 cycles                                                 (83.05%)
363      42518139 instructions    #   1.06 insns per cycle               (83.39%)
364           123 llc_misses      #   2.89 LLC misses per million insns  (83.15%)
365
366|
367| Output below is for the trace logs.
368| Run in separate terminals:
369| **# bpftool prog tracelog**
370| **# bpftool prog load -L -d file.o**
371
372::
373
374    bpftool-620059  [004] d... 2634685.517903: bpf_trace_printk: btf_load size 665 r=5
375    bpftool-620059  [004] d... 2634685.517912: bpf_trace_printk: map_create sample_map idx 0 type 2 value_size 4 value_btf_id 0 r=6
376    bpftool-620059  [004] d... 2634685.517997: bpf_trace_printk: prog_load sample insn_cnt 13 r=7
377    bpftool-620059  [004] d... 2634685.517999: bpf_trace_printk: close(5) = 0
378