[dpdk-dev] [PATCH v2] librte_cmdline: fix parsing initialisation

Olivier Matz olivier.matz at 6wind.com
Fri Jun 23 11:04:19 CEST 2017


Hi Adrien,

On Thu, 15 Jun 2017 12:15:48 +0200, Adrien Mazarguil <adrien.mazarguil at 6wind.com> wrote:
> From: "Bernard.Iremonger" <Bernard.iremonger at intel.com>
> 
> The dyn_tokens array is initialised at the beginning of the cmdline_parse
> function. However when the inst_num variable is incremented later in the
> function the dyn_tokens array is not reinitialised so the tokens from the
> previous command are used.
> 
> The solution is to initialise the dyn_tokens array in all while (inst)
> loops, before calling match_inst().
> 
> Fixes: 4fffc05a2b2c ("cmdline: support dynamic tokens")
> 
> CC: stable at dpdk.org
> Signed-off-by: Bernard Iremonger <bernard.iremonger at intel.com>
> Signed-off-by: Adrien Mazarguil <adrien.mazarguil at 6wind.com>
> 
> ---
> 
> Nice catch Bernard. This issue can be seen when implementing several
> flow-like commands in testpmd.
> 
> While testing your original patch though, it appeared that it did not fully
> address the issue, as completion remained partially broken. Actually all
> match_inst() calls should be preceded by a memset(), so here's an updated
> version instead of requesting you to do these changes.
> ---
>  lib/librte_cmdline/cmdline_parse.c | 5 +++--
>  1 file changed, 3 insertions(+), 2 deletions(-)
> 
> diff --git a/lib/librte_cmdline/cmdline_parse.c b/lib/librte_cmdline/cmdline_parse.c
> index b814880..c1d9f23 100644
> --- a/lib/librte_cmdline/cmdline_parse.c
> +++ b/lib/librte_cmdline/cmdline_parse.c
> @@ -276,7 +276,6 @@ cmdline_parse(struct cmdline *cl, const char * buf)
>  		return CMDLINE_PARSE_BAD_ARGS;
>  
>  	ctx = cl->ctx;
> -	memset(&dyn_tokens, 0, sizeof(dyn_tokens));
>  
>  	/*
>  	 * - look if the buffer contains at least one line
> @@ -321,6 +320,7 @@ cmdline_parse(struct cmdline *cl, const char * buf)
>  		debug_printf("INST %d\n", inst_num);
>  
>  		/* fully parsed */
> +		memset(&dyn_tokens, 0, sizeof(dyn_tokens));
>  		tok = match_inst(inst, buf, 0, tmp_result.buf,
>  				 sizeof(tmp_result.buf), &dyn_tokens);
>  
> @@ -400,7 +400,6 @@ cmdline_complete(struct cmdline *cl, const char *buf, int *state,
>  
>  	debug_printf("%s called\n", __func__);
>  	memset(&token_hdr, 0, sizeof(token_hdr));
> -	memset(&dyn_tokens, 0, sizeof(dyn_tokens));
>  
>  	/* count the number of complete token to parse */
>  	for (i=0 ; buf[i] ; i++) {
> @@ -423,6 +422,7 @@ cmdline_complete(struct cmdline *cl, const char *buf, int *state,
>  		inst = ctx[inst_num];
>  		while (inst) {
>  			/* parse the first tokens of the inst */
> +			memset(&dyn_tokens, 0, sizeof(dyn_tokens));
>  			if (nb_token &&
>  			    match_inst(inst, buf, nb_token, NULL, 0,
>  				       &dyn_tokens))
> @@ -530,6 +530,7 @@ cmdline_complete(struct cmdline *cl, const char *buf, int *state,
>  		/* we need to redo it */
>  		inst = ctx[inst_num];
>  
> +		memset(&dyn_tokens, 0, sizeof(dyn_tokens));
>  		if (nb_token &&
>  		    match_inst(inst, buf, nb_token, NULL, 0, &dyn_tokens))
>  			goto next2;

It looks you call memset() before each call to match_inst(). So, I
have 2 questions:

- why not putting this memset() at the beginning of match_inst()?

- does the following test at the beginning of match_inst() still make sense?

        if (!token_p && dyn_tokens && inst->f) {
                if (!(*dyn_tokens)[0])            /* <<<<<<<<< here */
                        inst->f(&(*dyn_tokens)[0], NULL, dyn_tokens);
                token_p = (*dyn_tokens)[0];
        }


Thanks,
Olivier



More information about the dev mailing list