From: David Howells <dhowells@redhat.com>
In-Reply-To: <20041123085608.3c30aa34.akpm@osdl.org> 
References: <20041123085608.3c30aa34.akpm@osdl.org>  <20041123081129.3e0121fd.akpm@osdl.org> <20041122155434.758c6fff.akpm@osdl.org> <11948.1101130077@redhat.com> <29356.1101201515@redhat.com> <15564.1101228539@redhat.com> 
To: Andrew Morton <akpm@osdl.org>
Cc: torvalds@osdl.org, wli@holomorphy.com, hch@infradead.org,
       linux-kernel@vger.kernel.org
Subject: Re: [PATCH] Compound page overhaul 
User-Agent: EMH/1.14.1 SEMI/1.14.5 (Awara-Onsen) FLIM/1.14.5 (Demachiyanagi) APEL/10.6 Emacs/21.3 (i386-redhat-linux-gnu) MULE/5.0 (SAKAKI)
MIME-Version: 1.0 (generated by SEMI 1.14.5 - "Awara-Onsen")
Content-Type: text/plain; charset=US-ASCII
Date: Tue, 23 Nov 2004 17:48:18 +0000
Message-ID: <16887.1101232098@redhat.com>
Sender: linux-kernel-owner@vger.kernel.org
Content-Length: 1636
Lines: 36


> So why did you create a "Compound page overhaul" in the first place?  Was it
> not to address some insufficiency for !MMU?

Not entirely. Part of it was to improve for !MMU use, and part of it was it
looked like I could improve it in general both by making it more readable and
by things such as making page->private available on the head page.

Linus suggested adding a CONFIG_COMPOUND_PAGE or something similar. By making
half of the compound page stuff mandatory I could also get rid of some
#ifdefs[*] for what appears to be a small overhead when allocating high-order
pages when HUGETLBFS is not defined by making use of the fact that we'd be
tickling the cache over the secondary page structures anyway.

[*] People seem to want to give me the impression that #ifdefs are evil and
    should all be buried at least 10 feet down:-)

This in turn provides a way to simplify a number of other things, such as the
"free_pages" functions.

There should be no overhead on single page handling when
ENHANCED_COMPOUND_PAGES is not set. If it is set, then the overhead is pretty
much the same as for hugetlbfs being compiled in now.

> The current compound page logic should handle that quite happily, no?

The current compound page implementation takes page->private away. What I've
done gives it back, currently at the cost of one page flag bit, but there are
ways around even that.

David
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.kernel.org
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/