Received-SPF: pass (google.com: best guess record for domain of linux-kernel-owner@vger.kernel.org designates 209.132.180.67 as permitted sender) client-ip=209.132.180.67;
MIME-Version: 1.0
References: <20180627013116.12411-1-bhe@redhat.com> <20180627013116.12411-5-bhe@redhat.com>
 <cb67381c-078c-62e6-e4c0-9ecf3de9e84d@intel.com>
In-Reply-To: <cb67381c-078c-62e6-e4c0-9ecf3de9e84d@intel.com>
From:   Pavel Tatashin <pasha.tatashin@oracle.com>
Date:   Fri, 29 Jun 2018 13:48:06 -0400
Message-ID: <CAGM2rebsL_fS8XKRvN34NWiFN3Hh63ZOD8jDj8qeSOUPXcZ2fA@mail.gmail.com>
Subject: Re: [PATCH v5 4/4] mm/sparse: Optimize memmap allocation during sparse_init()
To:     dave.hansen@intel.com
Cc:     bhe@redhat.com, LKML <linux-kernel@vger.kernel.org>,
        Andrew Morton <akpm@linux-foundation.org>, pagupta@redhat.com,
        Linux Memory Management List <linux-mm@kvack.org>,
        kirill.shutemov@linux.intel.com
Content-Type: text/plain; charset="UTF-8"
Sender: linux-kernel-owner@vger.kernel.org
Precedence: bulk

> > +      * increase 'nr_consumed_maps' whether its allocation of memmap
> > +      * or usemap failed or not, so that after we handle the i-th
> > +      * memory section, can get memmap and usemap of (i+1)-th section
> > +      * correctly. */
>
> This makes no sense to me.  Why are we incrementing 'nr_consumed_maps'
> when we do not consume one?
>
> You say that we increment it so that things will work, but not how or
> why it makes things work.  I'm confused.

Hi Dave,

nr_consumed_maps is a local counter. map_map contains struct pages for
each section. In order to assign them to correct sections this local
counter must be incremented even when some parts of map_map are empty.

Here is example:
Node1:
map_map[0] -> Struct pages ...
map_map[1] -> NULL
Node2:
map_map[2] -> Struct pages ...

We always want to configure section from Node2 with struct pages from
Node2. Even, if there are holes in-between. The same with usemap.

Pavel