Re: [PATCH 05/60] iwlagn: introduce struct iwl-shared - known by all layers

[wwguy <wey-yi.w.guy@intel.com>]

On Tue, 2011-08-16 at 02:26 -0700, Stanislaw Gruszka wrote:
> On Mon, Aug 15, 2011 at 08:29:12AM -0700, Guy, Wey-Yi wrote:
> > On Mon, 2011-08-15 at 08:12 -0700, Stanislaw Gruszka wrote:
> > > On Mon, Aug 15, 2011 at 07:10:22AM -0700, Guy, Wey-Yi wrote:
> > > > On Mon, 2011-08-15 at 07:44 -0700, Stanislaw Gruszka wrote:
> > > > > On Mon, Aug 15, 2011 at 06:46:45AM -0700, Guy, Wey-Yi wrote:
> > > > > > On Mon, 2011-08-15 at 07:21 -0700, Stanislaw Gruszka wrote:
> > > > > > > On Mon, Aug 15, 2011 at 06:13:39AM -0700, Guy, Wey-Yi wrote:
> > > > > > > > now we are working on separate iwlagn driver into two layers (upper and
> > > > > > > > lower), the lower layer contains both bus and transport sub-layer which
> > > > > > > > is hw dependent, and the upper layer should be bus/hw independent. by
> > > > > > > > doing so, the single driver can handle different bus and different core
> > > > > > > > architecture. Having the share data structure (priv->shrd->foo) between
> > > > > > > > two layers is for that reason (loosely couple).
> > > > > > > 
> > > > > > > You did not explained reason for introducing iwl_shared structure,
> > > > > > > actually you confirmed it is useless. You should simply share iwl_priv
> > > > > > > on any layer you have.
> > > > > > > 
> > > > > > > Stanislaw
> > > > > > 
> > > > > > It is not true, I mean "boring" but not "useless", it does not make
> > > > > > sense share all the priv with lower layer.
> > > > > 
> > > > > Why? Even if there are some fields that you will use only on
> > > > > upper layer, it does not mean that pointer to priv structure
> > > > > could not be passed to lower layer.
> > > > > 
> > > > why, if we doing that, lower layer need to know how to reference the
> > > > priv which I don't believe it is a good design. I want to make the upper
> > > > and lower layer as loose as possible and it is flexible for new
> > > > architecture.
> > > 
> > > Providing more abstraction layers than needed is worse.
> > 
> > I don't really see any issues with it
> 
> It make driver more complex than it could be.
> 
> >  and I don't believe it will cause
> > problem if we do it right. by doing so, we will improve the flexibility.
> 
> It does not offer any more flexibility than sharing iwl_priv, it
> is actually less flexible.
> 
how so, different upper layer can have differ private structure, just
has to include shrd which is shared with lower layer


> > > > > > think about I also mention
> > > > > > different core architecture, that mean we can have different upper layer
> > > > > > which has different priv; but shrd is always common to all different
> > > > > > upper layer to share with lower layer.
> > > > > 
> > > > > What different upper layer? Non mac80211, if so you should write
> > > > > separate driver for it.
> > > > 
> > > > not mac80211, it is still part of driver, but different upper layer
> > > > might/can contain different system flow. that is why I mention different
> > > > core architecture. now we are doing preparation work so you did not see
> > > > the new stuff yet.
> > > 
> > > Can we get opinion from Johannes about that, and eventually ACKs/NACKs
> > > for patches?
> > > 
> > 
> > it is pure iwlagn driver changes and not impact mac80211 at all.
> > Johannes already in the loop and he understand the changes.
> 
> I'm asking about Johannes opinion, not your opinion about his opinion :-)
> 
but I am not seeing any relation those patches related to mac80211

Wey