Re: [PATCH v1] config: make max memzones definition configurable

[Bruce Richardson <bruce.richardson@intel.com>]

On Mon, Feb 13, 2023 at 02:55:41PM +0100, Thomas Monjalon wrote:
> 13/02/2023 12:05, Bruce Richardson:
> > On Sun, Feb 12, 2023 at 10:53:19AM +0200, Ophir Munk wrote:
> > > In current DPDK the RTE_MAX_MEMZONE definition is unconditionally
> > > hard coded as 2560.  For applications requiring different values of
> > > this parameter – it is more convenient to set its value as part of
> > > the meson command line rather than changing the dpdk source code per
> > > application.  An example would be of an application that uses the
> > > DPDK mempool library which is based on DPDK memzone library.  The
> > > application may need to create a number of steering tables, each of
> > > which will require its own mempool allocation.  This commit adds a
> > > meson optional parameter named max_memzones. If not specified - it is
> > > set by default to 2560. The hard coded definition of RTE_MAX_MEMZONE
> > > is removed. During meson build time the RTE_MAX_MEMZONE can be
> > > optionally defined as the value of max_memzones parameter.
> > > 
> > > Signed-off-by: Ophir Munk <ophirmu@nvidia.com> --- RFC:
> > > https://patchwork.dpdk.org/project/dpdk/patch/20230130092302.376145-1-ophirmu@nvidia.com/
> > > 
> > >  config/meson.build  | 1 + config/rte_config.h | 1 -
> > >  meson_options.txt   | 2 ++ 3 files changed, 3 insertions(+), 1
> > >  deletion(-)
> > > 
> > Acked-by: Bruce Richardson <bruce.richardson@intel.com>
> 
> Are we going to move all compilation-defined settings to
> meson_options.txt?  The direction discussed in recent years was to
> configure things at runtime, and stop adding compilation-time settings.
> 
> In this case, it is quite easy to add a new function void
> rte_memzone_set_max(int max) to be called before rte_eal_init().  If not
> called, the historical default is used.
> 
Good point, I admit I had forgotten that.

Looking at the use of RTE_MAX_MEMZONE, it is used as an array dimension in
a number of places, but, from what I see on cursory examination, it should
be replacable with a runtime value without significant pain in most cases.
The one that probably needs more attention is the fact that the "net/qede"
driver maintains an array of memzones in it's base-code layer. Therefore,
we probably need input from that driver maintainer to know the impact there
and why that array is needed in a net driver. [Adding the two maintainers
on CC]

/Bruce