7. Command-line Library

Since its earliest versions, DPDK has included a command-line library - primarily for internal use by, for example, dpdk-testpmd and the dpdk-test binaries, but the library is also exported on install and can be used by any end application. This chapter covers the basics of the command-line library and how to use it in an application.

7.1. Library Features

The DPDK command-line library supports the following features:

  • Tab-completion available for interactive terminal sessions

  • Ability to read and process commands taken from an input file, e.g. startup script

  • Parameterized commands able to take multiple parameters with different datatypes:

    • Strings
    • Signed/unsigned 16/32/64-bit integers
    • IP Addresses
    • Ethernet Addresses
  • Ability to multiplex multiple commands to a single callback function

7.2. Adding Command-line to an Application

Adding a command-line instance to an application involves a number of coding steps.

  1. Define the result structure for the command, specifying the command parameters
  2. Provide an initializer for each field in the result
  3. Define the callback function for the command
  4. Provide a parse result structure instance for the command, linking the callback to the command
  5. Add the parse result structure to a command-line context
  6. Within your main application code, create a new command-line instance passing in the context.

Many of these steps can be automated using the script dpdk-cmdline-gen.py installed by DPDK, and found in the buildtools folder in the source tree. This section covers adding a command-line using this script to generate the boiler plate, while the following section, Worked Example of Adding Command-line to an Application covers the steps to do so manually.

7.2.1. Creating a Command List File

The dpdk-cmdline-gen.py script takes as input a list of commands to be used by the application. While these can be piped to it via standard input, using a list file is probably best.

The format of the list file must be:

  • Comment lines start with ‘#’ as first non-whitespace character
  • One command per line
  • Variable fields are prefixed by the type-name in angle-brackets, for example:
    • <STRING>message
    • <UINT16>port_id
    • <IP>src_ip
    • <IPv4>dst_ip4
    • <IPv6>dst_ip6
  • Variable fields, which take their values from a list of options, have the comma-separated option list placed in braces, rather than a the type name. For example,
    • <(rx,tx,rxtx)>mode
  • The help text for a command is given in the form of a comment on the same line as the command

An example list file, with a variety of (unrelated) commands, is shown below:

# example list file
list                     # show all entries
add <UINT16>x <UINT16>y  # add x and y
echo <STRING>message     # print message to screen
add socket <STRING>path  # add unix socket with the given path
set mode <(rx,tx)>rxtx   # set Rx-only or Tx-only mode
quit                     # close the application

7.2.2. Running the Generator Script

To generate the necessary definitions for a command-line, run dpdk-cmdline-gen.py passing the list file as parameter. The script will output the generated C code to standard output, the contents of which are in the form of a C header file. Optionally, an output filename may be specified via the -o/--output-file argument.

The generated content includes:

  • The result structure definitions for each command
  • The token initializers for each structure field
  • An “extern” function prototype for the callback for each command
  • A parse context for each command, including the per-command comments as help string
  • A command-line context array definition, suitable for passing to cmdline_new

If so desired, the script can also output function stubs for the callback functions for each command. This behaviour is triggered by passing the --stubs flag to the script. In this case, an output file must be provided with a filename ending in “.h”, and the callback stubs will be written to an equivalent “.c” file.

Note

The stubs are written to a separate file, to allow continuous use of the script to regenerate the command-line header, without overwriting any code the user has added to the callback functions. This makes it easy to incrementally add new commands to an existing application.

7.2.3. Providing the Function Callbacks

As discussed above, the script output is a header file, containing structure definitions, but the callback functions themselves obviously have to be provided by the user. These callback functions must be provided as non-static functions in a C file, and named cmd_<cmdname>_parsed. The function prototypes can be seen in the generated output header.

The “cmdname” part of the function name is built up by combining the non-variable initial tokens in the command. So, given the commands in our worked example below: quit and show port stats <n>, the callback functions would be:

void
cmd_quit_parsed(void *parsed_result, struct cmdline *cl, void *data)
{
     ...
}

void
cmd_show_port_stats_parsed(void *parsed_result, struct cmdline *cl, void *data)
{
     ...
}

These functions must be provided by the developer, but, as stated above, stub functions may be generated by the script automatically using the --stubs parameter.

The same “cmdname” stem is used in the naming of the generated structures too. To get at the results structure for each command above, the parsed_result parameter should be cast to struct cmd_quit_result or struct cmd_show_port_stats_result respectively.

Note

In some cases, the user may want to have an optional variable parameter at the end of a command. Such a variable parameter would not normally be included in the “cmdname” string, leading to duplicate definition errors. To work around this, any variable token with a name prefixed by '__' will be included in the “cmdname” string, with the prefix removed. Using this, it is possible to have commands, such as: start tx_first and start tx_first <UINT16>__n, without them conflicting. The resulting code generated will expect functions called cmd_start_tx_first_parsed and cmd_start_tx_first_n_parsed respectively.

7.2.4. Integrating with the Application

To integrate the script output with the application, we must #include the generated header into our applications C file, and then have the command-line created via either cmdline_new or cmdline_stdin_new. The first parameter to the function call should be the context array in the generated header file, ctx by default. (Modifiable via script parameter).

The callback functions may be in this same file, or in a separate one - they just need to be available to the linker at build-time.

7.2.5. Limitations of the Script Approach

The script approach works for most commands that a user may wish to add to an application. However, it does not support the full range of functions possible with the DPDK command-line library. For example, it is not possible using the script to multiplex multiple commands into a single callback function. To use this functionality, the user should follow the instructions in the next section Worked Example of Adding Command-line to an Application to manually configure a command-line instance.

7.3. Worked Example of Adding Command-line to an Application

The next few subsections will cover each of the steps listed in Adding Command-line to an Application in more detail, working through an example to add two commands to a command-line instance. Those two commands will be:

  1. quit - as the name suggests, to close the application
  2. show port stats <n> - to display on-screen the statistics for a given ethernet port

Note

For further examples of use of the command-line, see cmdline example application

7.3.1. Defining Command Result Structure

The first structure to be defined is the structure which will be created on successful parse of a command. This structure contains one member field for each token, or word, in the command. The simplest case is for a one-word command, like quit. For this, we only need to define a structure with a single string parameter to contain that word.

struct cmd_quit_result {
   cmdline_fixed_string_t quit;
};

For readability, the name of the struct member should match that of the token in the command.

For our second command, we need a structure with four member fields in it, as there are four words/tokens in our command. The first three are strings, and the final one is a 16-bit numeric value. The resulting struct looks like:

struct cmd_show_port_stats_result {
   cmdline_fixed_string_t show;
   cmdline_fixed_string_t port;
   cmdline_fixed_string_t stats;
   uint16_t n;
};

As before, we choose names to match the tokens in the command. Since our numeric parameter is a 16-bit value, we use uint16_t type for it. Any of the standard sized integer types can be used as parameters, depending on the desired result.

Beyond the standard integer types, the library also allows variable parameters to be of a number of other types, as called out in the feature list above.

  • For variable string parameters, the type should be cmdline_fixed_string_t - the same as for fixed tokens, but these will be initialized differently (as described below).
  • For ethernet addresses use type struct rte_ether_addr
  • For IP addresses use type cmdline_ipaddr_t

7.3.2. Providing Field Initializers

Each field of our result structure needs an initializer. For fixed string tokens, like “quit”, “show” and “port”, the initializer will be the string itself.

static cmdline_parse_token_string_t cmd_quit_quit_tok =
   TOKEN_STRING_INITIALIZER(struct cmd_quit_result, quit, "quit");

The convention for naming used here is to include the base name of the overall result structure - cmd_quit in this case, as well as the name of the field within that structure - quit in this case, followed by _tok. (This is why there is a double quit in the name above).

This naming convention is seen in our second example, which also demonstrates how to define a numeric initializer.

static cmdline_parse_token_string_t cmd_show_port_stats_show_tok =
   TOKEN_STRING_INITIALIZER(struct cmd_show_port_stats_result, show, "show");
static cmdline_parse_token_string_t cmd_show_port_stats_port_tok =
   TOKEN_STRING_INITIALIZER(struct cmd_show_port_stats_result, port, "port");
static cmdline_parse_token_string_t cmd_show_port_stats_stats_tok =
   TOKEN_STRING_INITIALIZER(struct cmd_show_port_stats_result, stats, "stats");
static cmdline_parse_token_num_t cmd_show_port_stats_n_tok =
   TOKEN_NUM_INITIALIZER(struct cmd_show_port_stats_result, n, RTE_UINT16);

For variable string tokens, the same TOKEN_STRING_INITIALIZER macro should be used. However, the final parameter should be NULL rather than a hard-coded token string.

For numeric parameters, the final parameter to the TOKEN_NUM_INITIALIZER macro should be the cmdline type matching the variable type defined in the result structure, e.g. RTE_UINT8, RTE_UINT32, etc.

For IP addresses, the macro TOKEN_IPADDR_INITIALIZER should be used.

For ethernet addresses, the macro TOKEN_ETHERADDR_INITIALIZER should be used.

7.3.3. Defining Callback Function

For each command, we need to define a function to be called once the command has been recognised. The callback function should have type:

void (*f)(void *, struct cmdline *, void *)

where the first parameter is a pointer to the result structure defined above, the second parameter is the command-line instance, and the final parameter is a user-defined pointer provided when we associate the callback with the command. Most callback functions only use the first parameter, or none at all, but the additional two parameters provide some extra flexibility, to allow the callback to work with non-global state in your application.

For our two example commands, the relevant callback functions would look very similar in definition. However, within the function body, we assume that the user would need to reference the result structure to extract the port number in the second case.

void
cmd_quit_parsed(void *parsed_result, struct cmdline *cl, void *data)
{
   quit = 1;
}
void
cmd_show_port_stats_parsed(void *parsed_result, struct cmdline *cl, void *data)
{
   struct cmd_show_port_stats_result *res = parsed_result;
   uint16_t port_id = res->n;
   ...
}

7.3.4. Associating Callback and Command

The cmdline_parse_inst_t type defines a “parse instance”, i.e. a sequence of tokens to be matched and then an associated function to be called. Also included in the instance type are a field for help text for the command, and any additional user-defined parameter to be passed to the callback functions referenced above. For example, for our simple “quit” command:

static cmdline_parse_inst_t cmd_quit = {
    .f = cmd_quit_parsed,
    .data = NULL,
    .help_str = "Close the application",
    .tokens = {
        (void *)&cmd_quit_quit_tok,
        NULL
    }
};

In this case, we firstly identify the callback function to be called, then set the user-defined parameter to NULL, provide a help message to be given, on request, to the user explaining the command, before finally listing out the single token to be matched for this command instance.

For our second, port stats, example, as well as making things a little more complicated by having multiple tokens to be matched, we can also demonstrate passing in a parameter to the function. Let us suppose that our application does not always use all the ports available to it, but instead only uses a subset of the ports, stored in an array called active_ports. Our stats command, therefore, should only display stats for the currently in-use ports, so we pass this active_ports array. (For simplicity of illustration, we shall assume that the array uses a terminating marker, e.g. -1 for the end of the port list, so we don’t need to pass in a length parameter too.)

extern int16_t active_ports[];
...
static cmdline_parse_inst_t cmd_show_port_stats = {
    .f = cmd_show_port_stats_parsed,
    .data = active_ports,
    .help_str = "Show statistics for active network ports",
    .tokens = {
        (void *)&cmd_show_port_stats_show_tok,
        (void *)&cmd_show_port_stats_port_tok,
        (void *)&cmd_show_port_stats_stats_tok,
        (void *)&cmd_show_port_stats_n_tok,
        NULL
    }
};

7.3.5. Adding Command to Command-line Context

Now that we have configured each individual command and callback, we need to merge these into a single array of command-line “contexts”. This context array will be used to create the actual command-line instance in the application. Thankfully, each context entry is the same as each parse instance, so our array is defined by simply listing out the previously defined command parse instances.

static cmdline_parse_ctx_t ctx[] = {
    &cmd_quit,
    &cmd_show_port_stats,
    NULL
};

The context list must be terminated by a NULL entry.

7.3.6. Creating a Command-line Instance

Once we have our ctx variable defined, we now just need to call the API to create the new command-line instance in our application. The basic API is cmdline_new which will create an interactive command-line with all commands available. However, if additional features for interactive use - such as tab-completion - are desired, it is recommended that cmdline_new_stdin be used instead.

A pattern that can be used in applications is to use cmdline_new for processing any startup commands, either from file or from the environment (as is done in the “dpdk-test” application), and then using cmdline_stdin_new thereafter to handle the interactive part. For example, to handle a startup file and then provide an interactive prompt:

struct cmdline *cl;
int fd = open(startup_file, O_RDONLY);

if (fd >= 0) {
    cl = cmdline_new(ctx, "", fd, STDOUT_FILENO);
    if (cl == NULL) {
        /* error handling */
    }
    cmdline_interact(cl);
    cmdline_quit(cl);
    close(fd);
}

cl = cmdline_stdin_new(ctx, "Proxy>> ");
if (cl == NULL) {
    /* error handling */
}
cmdline_interact(cl);
cmdline_stdin_exit(cl);

7.3.7. Multiplexing Multiple Commands to a Single Function

To reduce the amount of boiler-plate code needed when creating a command-line for an application, it is possible to merge a number of commands together to have them call a separate function. This can be done in a number of different ways:

  • A callback function can be used as the target for a number of different commands. Which command was used for entry to the function can be determined by examining the first parameter, parsed_result in our examples above.
  • For simple string commands, multiple options can be concatenated using the “#” character. For example: exit#quit, specified as a token initializer, will match either on the string “exit” or the string “quit”.

As a concrete example, these two techniques are used in the DPDK unit test application dpdk-test, where a single command cmdline_parse_t instance is used for all the “dump_<item>” test cases.

static void cmd_dump_parsed(void *parsed_result,
			    __rte_unused struct cmdline *cl,
			    __rte_unused void *data)
{
	struct cmd_dump_result *res = parsed_result;

	if (!strcmp(res->dump, "dump_physmem"))
		rte_dump_physmem_layout(stdout);
	else if (!strcmp(res->dump, "dump_memzone"))
		rte_memzone_dump(stdout);
	else if (!strcmp(res->dump, "dump_struct_sizes"))
		dump_struct_sizes();
	else if (!strcmp(res->dump, "dump_ring"))
		rte_ring_list_dump(stdout);
	else if (!strcmp(res->dump, "dump_mempool"))
		rte_mempool_list_dump(stdout);
	else if (!strcmp(res->dump, "dump_devargs"))
		rte_devargs_dump(stdout);
	else if (!strcmp(res->dump, "dump_log_types"))
		rte_log_dump(stdout);
	else if (!strcmp(res->dump, "dump_malloc_stats"))
		rte_malloc_dump_stats(stdout, NULL);
	else if (!strcmp(res->dump, "dump_malloc_heaps"))
		rte_malloc_dump_heaps(stdout);
}

cmdline_parse_token_string_t cmd_dump_dump =
	TOKEN_STRING_INITIALIZER(struct cmd_dump_result, dump,
				 "dump_physmem#"
				 "dump_memzone#"
				 "dump_struct_sizes#"
				 "dump_ring#"
				 "dump_mempool#"
				 "dump_malloc_stats#"
				 "dump_malloc_heaps#"
				 "dump_devargs#"
				 "dump_log_types");

cmdline_parse_inst_t cmd_dump = {
	.f = cmd_dump_parsed,  /* function to call */
	.data = NULL,      /* 2nd arg of func */
	.help_str = "dump status",
	.tokens = {        /* token list, NULL terminated */
		(void *)&cmd_dump_dump,
		NULL,
	},
};

7.4. Examples of Command-line Use in DPDK

To help the user follow the steps provided above, the following DPDK files can be consulted for examples of command-line use.

Note

This is not an exhaustive list of examples of command-line use in DPDK. It is simply a list of a few files that may be of use to the application developer. Some of these referenced files contain more complex examples of use that others.

  • commands.c/.h in examples/cmdline
  • mp_commands.c/.h in examples/multi_process/simple_mp
  • commands.c in app/test