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