Re: [RFC v2] genalloc:add an gen_pool_first_fit_align algo to genalloc

[Scott Wood <scottwood@freescale.com>]

On Wed, 2015-07-29 at 20:27 -0500, Zhao Qiang-B45475 wrote:
> On Thu, 2015-07-30 at 5:21, Scott Wood wrote:
> > -----Original Message-----
> > From: Wood Scott-B07421
> > Sent: Thursday, July 30, 2015 12:19 AM
> > To: Zhao Qiang-B45475
> > Cc: lauraa@codeaurora.org; linux-kernel@vger.kernel.org; linuxppc-
> > dev@lists.ozlabs.org; akpm@linux-foundation.org; olof@lixom.net;
> > catalin.marinas@arm.com; Xie Xiaobo-R63061
> > Subject: Re: [RFC v2] genalloc:add an gen_pool_first_fit_align algo to
> > genalloc
> > 
> > On Tue, 2015-07-28 at 00:32 -0500, Zhao Qiang-B45475 wrote:
> > > On Tue, 2015-07-28 at 5:21, Scott Wood wrote:
> > > > -----Original Message-----
> > > > From: Wood Scott-B07421
> > > > Sent: Tuesday, July 28, 2015 5:21 AM
> > > > To: Zhao Qiang-B45475
> > > > Cc: lauraa@codeaurora.org; linux-kernel@vger.kernel.org; linuxppc-
> > > > dev@lists.ozlabs.org; akpm@linux-foundation.org; olof@lixom.net;
> > > > catalin.marinas@arm.com; Xie Xiaobo-R63061
> > > > Subject: Re: [RFC v2] genalloc:add an gen_pool_first_fit_align algo
> > > > to genalloc
> > > > 
> > > > On Mon, 2015-07-27 at 17:57 +0800, Zhao Qiang wrote:
> > > > 
> > > > Where's the part that adds the ability to pass in data to each
> > > > allocation call, as per the previous discussion?
> > > 
> > > You means to use gen_pool_alloc_data()?
> > 
> > Yes.
> > 
> > > Previously you said that the format of data is algorithm-specific, So
> > > I think it is better to handle data in algorithm function.
> > 
> > It is a channel for communication from the API caller to the algorithm.
> > 
> > > If you still prefer gen_pool_alloc_data(), I will modify it.
> > > But there still are details I want to confirm with you.
> > > 1. If use gen_pool_alloc_data(), should I pass data as a parameter?
> > 
> > Yes.
> > 
> > > 2. Should I count align_mask in gen_pool_alloc_data(), meanwhile, add
> > >    a align_mask to data as a member?
> > 
> > gen_pool_alloc_data() should just pass data to the algorithm.  The
> > algorithm should calculate align_mask based on align.  I don't think
> > exposing align_mask to API users would be very friendly.
> 
> If calculate align_mask in algorithm, I need to get pool->min_alloc_order 
> in algorithm,
> Like:
>                order = data->pool->min_alloc_order;
>                align_mask = ((data->align + (1UL << order) - 1) >> order) - 
> 1; 
> so I add pool to structure data as a member. Is there any other better 
> idea? 

Pass pool as a parameter to the algorithm.

-Scott