Re: [PATCH v3 3/7] service cores: coremask parsing

["Van Haaren, Harry" <harry.van.haaren@intel.com>]

> From: Jerin Jacob [mailto:jerin.jacob@caviumnetworks.com]
> Sent: Tuesday, July 4, 2017 1:46 PM
> To: Van Haaren, Harry <harry.van.haaren@intel.com>
> Cc: dev@dpdk.org; thomas@monjalon.net; Wiles, Keith <keith.wiles@intel.com>; Richardson,
> Bruce <bruce.richardson@intel.com>
> Subject: Re: [PATCH v3 3/7] service cores: coremask parsing
> 
> -----Original Message-----
> > Date: Sun, 2 Jul 2017 22:35:10 +0100
> > From: Harry van Haaren <harry.van.haaren@intel.com>
> > To: dev@dpdk.org
> > CC: jerin.jacob@caviumnetworks.com, thomas@monjalon.net,
> >  keith.wiles@intel.com, bruce.richardson@intel.com, Harry van Haaren
> >  <harry.van.haaren@intel.com>
> > Subject: [PATCH v3 3/7] service cores: coremask parsing
> > X-Mailer: git-send-email 2.7.4
> >
> > Add logic for parsing a coremask from EAL, which allows
> > the application to be unaware of the cores being taken from
> > its coremask.
> >
> > Signed-off-by: Harry van Haaren <harry.van.haaren@intel.com>
> >
> > diff --git a/lib/librte_eal/common/eal_common_options.c
> b/lib/librte_eal/common/eal_common_options.c
> > index f470195..cee200c 100644
> > --- a/lib/librte_eal/common/eal_common_options.c
> > +++ b/lib/librte_eal/common/eal_common_options.c
> > @@ -61,6 +61,7 @@ const char
> >  eal_short_options[] =
> >  	"b:" /* pci-blacklist */
> >  	"c:" /* coremask */
> > +	"s:" /* service coremask */
> >  	"d:" /* driver */
> >  	"h"  /* help */
> >  	"l:" /* corelist */
> > @@ -267,6 +268,73 @@ static int xdigit2val(unsigned char c)
> >  }
> 
> Missing the --help update for service coremask details.
> 
> I think, EAL arguments are documented in another area of doc directory
> as well. Update the documents.

Will double check / fix this. Replying here now to advance discussion below; 

> >  static int
> > +eal_parse_service_coremask(const char *coremask)
> > +{
> > +	struct rte_config *cfg = rte_eal_get_configuration();
> > +	int i, j, idx = 0;
> > +	unsigned int count = 0;
> > +	char c;
> > +	int val;
> > +
> > +	if (coremask == NULL)
> > +		return -1;
> > +	/* Remove all blank characters ahead and after .
> > +	 * Remove 0x/0X if exists.
> > +	 */
> > +	while (isblank(*coremask))
> > +		coremask++;
> > +	if (coremask[0] == '0' && ((coremask[1] == 'x')
> > +		|| (coremask[1] == 'X')))
> > +		coremask += 2;
> > +	i = strlen(coremask);
> > +	while ((i > 0) && isblank(coremask[i - 1]))
> > +		i--;
> > +
> > +	if (i == 0)
> > +		return -1;
> > +
> > +	for (i = i - 1; i >= 0 && idx < RTE_MAX_LCORE; i--) {
> > +		c = coremask[i];
> > +		if (isxdigit(c) == 0) {
> > +			/* invalid characters */
> > +			return -1;
> > +		}
> > +		val = xdigit2val(c);
> > +		for (j = 0; j < BITS_PER_HEX && idx < RTE_MAX_LCORE;
> > +				j++, idx++) {
> > +			if ((1 << j) & val) {
> > +				/* handle master lcore already parsed */
> > +				uint32_t lcore = idx;
> > +				if (master_lcore_parsed &&
> > +						cfg->master_lcore == lcore)
> > +					continue;
> > +
> > +				if (!lcore_config[idx].detected) {
> > +					RTE_LOG(ERR, EAL,
> > +						"lcore %u unavailable\n", idx);
> > +					return -1;
> > +				}
> > +				lcore_config[idx].core_role = ROLE_SERVICE;
> 
> Why not to use rte_service_lcore_add(idx) here. So that in future some
> changes we don't need to touch this file.

The issue here is that the hugepages memory that service-cores requires is not available at this point. Keep in mind, the EAL parse-opts runs before almost anything else (makes sense, given we can specify e.g. --no-huge).

Given that there is not rte_malloc() available at this point, we have a few options:
1) Use existing allocated mem, e.g. the lcore_config[] array as above.
2) Delay the parsing of service-core mask until later. Breaks "parse -> validate-> config -> run" workflow.
3) Allocate temp memory to store the service-core indexes, and later free that back (feels hacky to me?)

Current scheme of (1) makes the most sense to me.


> I added following code in unit testcase and I have 8 cores system. So I
> was expecting cores prints from "0 3 4 5 6 7" as lcore 1 and 2 will be
> stolen by service core. But it looks like RTE_LCORE_FOREACH not honoring
> previous rte_service_lcore_add() functions.
> 
> testsuite_setup(void)
> {
> +       int i;
> +       rte_service_lcore_add(1);
> +       rte_service_lcore_add(2);
> +
> +       RTE_LCORE_FOREACH(i)
> +               printf("cores %d\n", i);


Root cause found - and fixed. If you don't strongly object to lcore_config[] method above, then I can prioritize this and try get a patchset up ASAP.