Re: [RFC 0/6] Regressions for "imply" behavior change

From: Arnd Bergmann <arnd@arndb.de>
To: Saeed Mahameed <saeedm@mellanox.com>
Cc: "narmstrong@baylibre.com" <narmstrong@baylibre.com>,
	"masahiroy@kernel.org" <masahiroy@kernel.org>,
	"Laurent.pinchart@ideasonboard.com" 
	<Laurent.pinchart@ideasonboard.com>,
	"leon@kernel.org" <leon@kernel.org>,
	"davem@davemloft.net" <davem@davemloft.net>,
	"linux-renesas-soc@vger.kernel.org" 
	<linux-renesas-soc@vger.kernel.org>,
	"linux-kernel@vger.kernel.org" <linux-kernel@vger.kernel.org>,
	"nico@fluxnic.net" <nico@fluxnic.net>,
	"linux-rdma@vger.kernel.org" <linux-rdma@vger.kernel.org>,
	"dri-devel@lists.freedesktop.org"
	<dri-devel@lists.freedesktop.org>,
	"kieran.bingham+renesas@ideasonboard.com" 
	<kieran.bingham+renesas@ideasonboard.com>,
	"jani.nikula@linux.intel.com" <jani.nikula@linux.intel.com>,
	"a.hajda@samsung.com" <a.hajda@samsung.com>,
	"jonas@kwiboo.se" <jonas@kwiboo.se>,
	"netdev@vger.kernel.org" <netdev@vger.kernel.org>,
	"airlied@linux.ie" <airlied@linux.ie>,
	"jgg@ziepe.ca" <jgg@ziepe.ca>,
	"jernej.skrabec@siol.net" <jernej.skrabec@siol.net>
Subject: Re: [RFC 0/6] Regressions for "imply" behavior change
Date: Thu, 16 Apr 2020 09:20:40 +0200	[thread overview]
Message-ID: <CAK8P3a3Wx5_bUOKnN3_hG5nLOqv3WCUtMSq6vOkJzWZgsmAz+A@mail.gmail.com> (raw)
In-Reply-To: <834c7606743424c64951dd2193ca15e29799bf18.camel@mellanox.com>

On Thu, Apr 16, 2020 at 5:25 AM Saeed Mahameed <saeedm@mellanox.com> wrote:
>
> On Tue, 2020-04-14 at 20:47 +0200, Arnd Bergmann wrote:
> > On Tue, Apr 14, 2020 at 7:49 PM Saeed Mahameed <saeedm@mellanox.com>
> > wrote:
> > > On Tue, 2020-04-14 at 17:25 +0200, Arnd Bergmann wrote:
> > > > On Tue, Apr 14, 2020 at 5:23 PM Jason Gunthorpe <jgg@ziepe.ca>
> > > > wrote:
> > > > Correct.
> > > >
> > >
> > > Great !
> > >
> > > Then bottom line we will change mlx5/Kconfig: to
> > >
> > > depends on VXLAN || !VXLAN
> >
> > Ok
> >
>
> BTW how about adding a new Kconfig option to hide the details of
> ( BAR || !BAR) ? as Jason already explained and suggested, this will
> make it easier for the users and developers to understand the actual
> meaning behind this tristate weird condition.
>
> e.g have a new keyword:
>      reach VXLAN
> which will be equivalent to:
>      depends on VXLAN && !VXLAN

I'd love to see that, but I'm not sure what keyword is best. For your
suggestion of "reach", that would probably do the job, but I'm not
sure if this ends up being more or less confusing than what we have
today.

> > > This will force MLX5_CORE to m when necessary to make vxlan
> > > reachable
> > > to mlx5_core.  So no need for explicit use of IS_REACHABLE().
> > > in mlx5 there are 4 of these:
> > >
> > >         imply PTP_1588_CLOCK
> > >         imply VXLAN
> > >         imply MLXFW
> > >         imply PCI_HYPERV_INTERFACE
> >
> > As mentioned earlier, we do need to replace the 'imply
> > PTP_1588_CLOCK'
> > with the same
> >
> >          depends on PTP_1588_CLOCK || !PTP_1588_CLOCK
> >
> > So far I have not seen problems for the other two options, so I
> > assume they
> > are fine for now -- it seems to build just fine without
> > PCI_HYPERV_INTERFACE,
> > and MLXFW has no other dependencies, meaning that 'imply' is the
> > same as 'select' here. Using 'select MLXFW' would make it clearer
> > perhaps.
>
> No, I would like to avoid select and allow building mlx5 without MLXFW,
> MLXFW already has a stub protected with IS_REACHABLE(), this is why we
> don't have an issue with it.

So the 'imply MLXFW' should be dropped then?

        Arnd