From mboxrd@z Thu Jan 1 00:00:00 1970 From: David Vrabel Subject: Re: [PATCH v4 0/2] make ballooned out pages have a valid mapping at all times Date: Mon, 5 Aug 2013 15:53:09 +0100 Message-ID: <51FFBC55.2090908@citrix.com> References: Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: List-Unsubscribe: , List-Post: List-Help: List-Subscribe: , Sender: xen-devel-bounces@lists.xen.org Errors-To: xen-devel-bounces@lists.xen.org To: Stefano Stabellini Cc: dcrisan@flexiant.com, xen-devel@lists.xensource.com, Ian Campbell , alex@alex.org.uk List-Id: xen-devel@lists.xenproject.org On 04/08/13 15:38, Stefano Stabellini wrote: > Hi all, > this patch series limits problems caused by tcp retransmits on NFS when > the original block pages were mapped from a foreign domain and now the > mapping is gone. > > It accomplishes the goal by: > > 1) mapping all ballooned out pages to a per-cpu "balloon_scratch_page"; > 2) making sure that once a grant is unmapped, the original mapping to > the per-cpu balloon_scratch_page is restored atomically. > > The first patch accomplishes (1), the second patch uses > GNTTABOP_unmap_and_replace to atomically unmap a grant and restore the > original mapping. > > > > Changes in this version: > - add an early_initcall to clear all the possible per_cpu > balloon_scratch_page. I know Konrad asked for this but I don't think it is necessary. Many other users of DEFINE_PER_CPU() assume the space is initialized to zero. e.g., cpufreq_show_table in drivers/cpufreq/freq_table.c If this is the only change in v4 I would suggest taking the v3 patches instead. David