From mboxrd@z Thu Jan  1 00:00:00 1970
From: Christoph Lameter <cl@linux-foundation.org>
Subject: Re: [PATCH 06/30] mm: kmem_alloc_estimate()
Date: Wed, 30 Jul 2008 15:02:07 -0500
Message-ID: <4890C8BF.4050908@linux-foundation.org>
References: <20080724140042.408642539@chello.nl>	 <20080724141529.716339226@chello.nl>	 <1217420503.7813.170.camel@penberg-laptop> <1217424662.8157.58.camel@twins>
Mime-Version: 1.0
Content-Type: text/plain; charset=ISO-8859-1
Content-Transfer-Encoding: 7bit
Cc: Pekka Enberg <penberg@cs.helsinki.fi>,
	Linus Torvalds <torvalds@linux-foundation.org>,
	Andrew Morton <akpm@linux-foundation.org>,
	linux-kernel@vger.kernel.org, linux-mm@kvack.org,
	netdev@vger.kernel.org, trond.myklebust@fys.uio.no,
	Daniel Lezcano <dlezcano@fr.ibm.com>,
	Neil Brown <neilb@suse.de>
To: Peter Zijlstra <a.p.zijlstra@chello.nl>
Return-path: <netdev-owner@vger.kernel.org>
Received: from smtp1.linux-foundation.org ([140.211.169.13]:48431 "EHLO
	smtp1.linux-foundation.org" rhost-flags-OK-OK-OK-OK)
	by vger.kernel.org with ESMTP id S1751808AbYG3UDM (ORCPT
	<rfc822;netdev@vger.kernel.org>); Wed, 30 Jul 2008 16:03:12 -0400
In-Reply-To: <1217424662.8157.58.camel@twins>
Sender: netdev-owner@vger.kernel.org
List-ID: <netdev.vger.kernel.org>

Peter Zijlstra wrote:

>>> +/*
>>> + * Calculate the upper bound of pages requires to sequentially allocate @bytes
>>> + * from kmalloc in an unspecified number of allocations of nonuniform size.
>>> + */
>>> +unsigned kmalloc_estimate_variable(gfp_t flags, size_t bytes)
>>> +{
>>> +	int i;
>>> +	unsigned long pages;
>>> +
>>> +	/*
>>> +	 * multiply by two, in order to account the worst case slack space
>>> +	 * due to the power-of-two allocation sizes.
>>> +	 */
>>> +	pages = DIV_ROUND_UP(2 * bytes, PAGE_SIZE);
>> For bytes > PAGE_SIZE this doesn't look right (for SLUB). We do page
>> allocator pass-through which means that we'll be grabbing high order
>> pages which can be bigger than what 'pages' is here.
> 
> Satisfying allocations from a bucket distribution with power-of-two
> (which page alloc order satisfies) has a worst case slack space of:
> 
> S(x) = 2^n - (2^(n-1)) - 1, n = ceil(log2(x))
> 
> This can be seen for the cases where x = 2^i + 1.

The needed bytes for a kmalloc allocation with size > PAGE_SIZE is

get_order(size) << PAGE_SHIFT bytes.

See kmalloc_large().