Re: [dpdk-dev] [PATCH v3 13/14] eal: unify internal config initialization

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

On Thu, Jul 4, 2019 at 12:50 PM Burakov, Anatoly <anatoly.burakov@intel.com>
wrote:

> On 04-Jul-19 8:56 AM, David Marchand wrote:
> > On Thu, Jul 4, 2019 at 9:50 AM David Marchand <david.marchand@redhat.com
> >
> > wrote:
> >
> >>
> >>
> >> On Thu, Jun 27, 2019 at 1:39 PM Anatoly Burakov <
> anatoly.burakov@intel.com>
> >> wrote:
> >>
> >>> Currently, each EAL will update internal/shared config in their
> >>> own way at init, resulting in needless duplication of code and
> >>> OS-dependent behavior. Move the functions to a common file and
> >>> add missing FreeBSD steps.
> >>>
> >>> Signed-off-by: Anatoly Burakov <anatoly.burakov@intel.com>
> >>> ---
> >>>   lib/librte_eal/common/eal_common_mcfg.c | 18 ++++++++++++++++++
> >>>   lib/librte_eal/common/eal_memcfg.h      |  8 ++++++++
> >>>   lib/librte_eal/freebsd/eal/eal.c        |  2 ++
> >>>   lib/librte_eal/linux/eal/eal.c          | 22 ++--------------------
> >>>   4 files changed, 30 insertions(+), 20 deletions(-)
> >>>
> >>> diff --git a/lib/librte_eal/common/eal_common_mcfg.c
> >>> b/lib/librte_eal/common/eal_common_mcfg.c
> >>> index dc6665d6a..fe8d2b726 100644
> >>> --- a/lib/librte_eal/common/eal_common_mcfg.c
> >>> +++ b/lib/librte_eal/common/eal_common_mcfg.c
> >>> @@ -31,6 +31,24 @@ eal_mcfg_wait_complete(void)
> >>>                  rte_pause();
> >>>   }
> >>>
> >>> +void
> >>> +eal_mcfg_update_internal(void)
> >>> +{
> >>> +       struct rte_mem_config *mcfg =
> >>> rte_eal_get_configuration()->mem_config;
> >>> +
> >>> +       internal_config.legacy_mem = mcfg->legacy_mem;
> >>> +       internal_config.single_file_segments =
> mcfg->single_file_segments;
> >>> +}
> >>> +
> >>> +void
> >>> +eal_mcfg_update_from_internal(void)
> >>> +{
> >>> +       struct rte_mem_config *mcfg =
> >>> rte_eal_get_configuration()->mem_config;
> >>> +
> >>> +       mcfg->legacy_mem = internal_config.legacy_mem;
> >>> +       mcfg->single_file_segments =
> internal_config.single_file_segments;
> >>> +}
> >>> +
> >>>   void
> >>>   rte_mcfg_mem_read_lock(void)
> >>>   {
> >>> diff --git a/lib/librte_eal/common/eal_memcfg.h
> >>> b/lib/librte_eal/common/eal_memcfg.h
> >>> index a2434417e..d02ac1621 100644
> >>> --- a/lib/librte_eal/common/eal_memcfg.h
> >>> +++ b/lib/librte_eal/common/eal_memcfg.h
> >>> @@ -68,6 +68,14 @@ struct rte_mem_config {
> >>>          uint8_t dma_maskbits;
> >>>   };
> >>>
> >>> +/* update internal config from shared mem config */
> >>> +void
> >>> +eal_mcfg_update_internal(void);
> >>> +
> >>> +/* update shared mem config from internal config */
> >>> +void
> >>> +eal_mcfg_update_from_internal(void);
> >>> +
> >>>   /* wait until primary process initialization is complete */
> >>>   void
> >>>   eal_mcfg_wait_complete(void);
> >>> diff --git a/lib/librte_eal/freebsd/eal/eal.c
> >>> b/lib/librte_eal/freebsd/eal/eal.c
> >>> index 13e230fc8..6bfe203fd 100644
> >>> --- a/lib/librte_eal/freebsd/eal/eal.c
> >>> +++ b/lib/librte_eal/freebsd/eal/eal.c
> >>> @@ -379,6 +379,7 @@ rte_config_init(void)
> >>>          case RTE_PROC_PRIMARY:
> >>>                  if (rte_eal_config_create() < 0)
> >>>                          return -1;
> >>> +               eal_mcfg_update_internal();
> >>>                  break;
> >>>          case RTE_PROC_SECONDARY:
> >>>                  if (rte_eal_config_attach() < 0)
> >>> @@ -386,6 +387,7 @@ rte_config_init(void)
> >>>                  eal_mcfg_wait_complete();
> >>>                  if (rte_eal_config_reattach() < 0)
> >>>                          return -1;
> >>> +               eal_mcfg_update_from_internal();
> >>>                  break;
> >>>          case RTE_PROC_AUTO:
> >>>          case RTE_PROC_INVALID:
> >>>
> >>
> >>
> >> Hum, you swapped eal_mcfg_update_internal and
> >> eal_mcfg_update_from_internal.
> >> The names are a bit ambiguous, and I wonder if we really need those
> >> separate helpers.
> >>
> >
> > Replying to myself... at least those helpers try to do something
> > explicitly: synchronising the local copy of the configuration with the
> > shared mem config.
> > Keeping them separate documents this step.
> >
> > Anyway, your choice, but you must fix freebsd :-)
> >
>
> Sorry, i don't follow - what needs to be fixed in FreeBSD?
>

eal_mcfg_update_from_internal() must be called in primary process.
eal_mcfg_update_internal() must be called in secondary processes.


-- 
David Marchand