Re: [PATCH 4/4] app/testpmd: enhance getopt_long usage

[David Marchand <david.marchand@redhat.com>]

On Tue, Mar 12, 2024 at 6:03 PM Ferruh Yigit <ferruh.yigit@amd.com> wrote:
>
> On 3/8/2024 2:48 PM, David Marchand wrote:
> > This is a cleanup similar to previous ones in EAL and examples.
> > Instead of using strcmp for every long options while getopt_long already
> > did such parsing, rely on getopt_long return value.
> >
> > Note for reviewers: this patch is best reviewed once applied locally and
> > displayed with git show -w.
> >
>
> Thanks for the cleanup, it was needed.

Thanks for the in depth review.

>
>
> > Signed-off-by: David Marchand <david.marchand@redhat.com>
> > ---
> >  app/test-pmd/parameters.c | 1928 +++++++++++++++++++++----------------
> >  1 file changed, 1084 insertions(+), 844 deletions(-)
> >
> > diff --git a/app/test-pmd/parameters.c b/app/test-pmd/parameters.c
> > index 271f0c995a..58f43cb5a8 100644
> > --- a/app/test-pmd/parameters.c
> > +++ b/app/test-pmd/parameters.c
> > @@ -40,6 +40,346 @@
> >
> >  #include "testpmd.h"
> >
> > +enum {
> > +     /* long options mapped to a short option */
> > +#define TESTPMD_OPT_AUTO_START "auto-start"
> > +     TESTPMD_OPT_AUTO_START_NUM = 'a',
> > +#define TESTPMD_OPT_HELP "help"
> > +     TESTPMD_OPT_HELP_NUM = 'h',
> > +#define TESTPMD_OPT_INTERACTIVE "interactive"
> > +     TESTPMD_OPT_INTERACTIVE_NUM = 'i',
> > +
> > +     /* first long only option value must be >= 256 */
> > +     TESTPMD_OPT_LONG_MIN_NUM = 256,
> >
>
> This is to be able to use all short options, not sure if this requires a
> comment.

Yes, those comments do not help.


>
> <...>
>
> > +};
> > +
> > +static struct option long_options[] = {
> >
>
> Can it be constant?

Yes.


>
> > +     { TESTPMD_OPT_AUTO_START, 0, NULL, TESTPMD_OPT_AUTO_START_NUM },
> >
>
> In original version "auto-start" long version is enabled only if
> 'RTE_LIB_CMDLINE' enabled, not sure why, but not for short version, I
> think there is a confusion.

I guess one reason why the --auto-start option is under
RTE_LIB_CMDLINE is because forwarding is always started when in non
interactive mode.
https://git.dpdk.org/dpdk/tree/app/test-pmd/testpmd.c#n4722

-a is available without RTE_LIB_CMDLINE, but it was probably not intentional.

In both cases, -a / --auto-start are a nop when testpmd is compiled
without RTE_LIB_CMDLINE.
So moving it out of the RTE_LIB_CMDLINE is not an issue.


>
> OK to remove RTE_LIB_CMDLINE requirement for 'auto-start' but perhaps
> better to have it in separate patch, what do you think?

Yes, it is worth a separate patch but see below my next comment about
RTE_LIB_CMDLINE check.

>
>
> > +     { TESTPMD_OPT_HELP, 0, NULL, TESTPMD_OPT_HELP_NUM },
> > +#ifdef RTE_LIB_CMDLINE
> > +     { TESTPMD_OPT_INTERACTIVE, 0, NULL, TESTPMD_OPT_INTERACTIVE_NUM },
> > +     { TESTPMD_OPT_CMDLINE_FILE, 1, NULL, TESTPMD_OPT_CMDLINE_FILE},
> >
>
> What about using 'no_argument' & 'required_argument' instead of '0' &
> '1', but I guess it will make lines too big,

Line length is the reason why I used the 0/1 but your suggestion looks
good to me.


> as flag will be always NULL in this approach, what about defining macros
> for arg and no_arg, like:
> ```
> #define ARG(name) (name), required_argument, NULL, name##_NUM
> #define NOARG(name) (name), no_argument, NULL, name##_NUM
> #define OPTARG(name) (name), optional_argument, NULL, name##_NUM
>
> { NOARG(TESTPMD_OPT_INTERACTIVE) },
> { ARG(TESTPMD_OPT_CMDLINE_FILE) },
> ```

It lgtm.


>
> > +     { TESTPMD_OPT_ETH_PEERS_CONFIGFILE, 1, NULL, TESTPMD_OPT_ETH_PEERS_CONFIGFILE_NUM },
> > +     { TESTPMD_OPT_ETH_PEER, 1, NULL, TESTPMD_OPT_ETH_PEER_NUM },
> >
>
> These long options are within "#ifdef RTE_LIB_CMDLINE" block, I don't
> know why, I guess above two can work without cmdline library.

Looking again at the options list, we have a nice mess in there... I
had not realised.

I think the only options that are conditional to RTE_LIB_CMDLINE should be:
-i, -a, --cmdline-file.

As I wrote above, leaving the -a/--auto-start option parsing out of
RTE_LIB_CMDLINE is not an issue.
So it would only leave -i/--interactive and --cmdline-file under
RTE_LIB_CMDLINE.
The rest can be moved out of RTE_LIB_CMDLINE.

And reading your next comment about RTE_LIB_CMDLINE, I think we are in sync.


>
> <...>
>
> > -
> >       argvopt = argv;
> >
>
> Why 'argvopt' is required, why not use 'argv' directly?

I'll have to double check.


>
> <...>
>
> >  #ifdef RTE_LIB_CMDLINE
> > ...
> > +             case TESTPMD_OPT_ETH_PEER_NUM: {
> > +                     char *port_end;
> >
> > -                             if (rte_ether_unformat_addr(port_end,
> > -                                             &peer_eth_addrs[n]) < 0)
> > -                                     rte_exit(EXIT_FAILURE,
> > -                                              "Invalid ethernet address: %s\n",
> > -                                              port_end);
> > -                             nb_peer_eth_addrs++;
> > -                     }
> > +                     errno = 0;
> > +                     n = strtoul(optarg, &port_end, 10);
> > +                     if (errno != 0 || port_end == optarg || *port_end++ != ',')
> > +                             rte_exit(EXIT_FAILURE,
> > +                                     "Invalid eth-peer: %s", optarg);
> > +                     if (n >= RTE_MAX_ETHPORTS)
> > +                             rte_exit(EXIT_FAILURE,
> > +                                     "eth-peer: port %d >= RTE_MAX_ETHPORTS(%d)\n",
> > +                                     n, RTE_MAX_ETHPORTS);
> > +
> > +                     if (rte_ether_unformat_addr(port_end,
> > +                                     &peer_eth_addrs[n]) < 0)
> > +                             rte_exit(EXIT_FAILURE,
> > +                                     "Invalid ethernet address: %s\n",
> > +                                     port_end);
> > +                     nb_peer_eth_addrs++;
> > +                     break;
> > +             }
> >  #endif
> >
>
> Similar to the 'struct option long_options' comment, I think expect from
> "interactive" & "cmdline-file", rest can be moved out of "#ifdef
> RTE_LIB_CMDLINE" block, but not this patch in different patch.

Yes.


>
> <...>
>
> >               default:
> >                       usage(argv[0]);
> >
>
> It is not shown in the patch, to print the error case it uses:
> `fprintf(stderr, "Invalid option: %s\n", argv[optind]);`
>
> But as far as I can see 'optind' is the next arg to call, so instead it
> should be 'optind - 1'.

Oh, good catch:

$ build-mini/app/dpdk-testpmd -c 3 --no-huge -m 40 -a 0:0.0 --vdev
net_null1 --vdev net_null2 -- -i --garbage
Invalid option: (null)
EAL: Error - exiting with code: 1
  Cause: Command line is incomplete or incorrect

>
>
> And just after while loop, there is "if (optind != argc)" check, I don't
> see when this check will be hit, so it looks useless.

But this part actually looks correct to me.
We could still hit this part with non handled trailing strings:

$ build-mini/app/dpdk-testpmd -c 3 --no-huge -m 40 -a 0:0.0 --vdev
net_null1 --vdev net_null2 -- -i garbage
Invalid parameter: garbage
EAL: Error - exiting with code: 1
  Cause: Command line is incorrect
Port 0 is closed
Port 1 is closed


-- 
David Marchand