[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <010001672d69716a-743e5098-eb2f-41ca-bf98-6e17eaaba00f-000000@email.amazonses.com>
Date: Mon, 19 Nov 2018 19:19:04 +0000
From: Christopher Lameter <cl@...ux.com>
To: Jerome Glisse <jglisse@...hat.com>
cc: Jason Gunthorpe <jgg@...pe.ca>, Leon Romanovsky <leon@...nel.org>,
Kenneth Lee <liguozhu@...ilicon.com>,
Tim Sell <timothy.sell@...sys.com>, linux-doc@...r.kernel.org,
Alexander Shishkin <alexander.shishkin@...ux.intel.com>,
Zaibo Xu <xuzaibo@...wei.com>, zhangfei.gao@...mail.com,
linuxarm@...wei.com, haojian.zhuang@...aro.org,
Hao Fang <fanghao11@...wei.com>,
Gavin Schenk <g.schenk@...elmann.de>,
RDMA mailing list <linux-rdma@...r.kernel.org>,
Zhou Wang <wangzhou1@...ilicon.com>,
Doug Ledford <dledford@...hat.com>,
Uwe Kleine-König
<u.kleine-koenig@...gutronix.de>,
David Kershner <david.kershner@...sys.com>,
Kenneth Lee <nek.in.cn@...il.com>,
Johan Hovold <johan@...nel.org>,
Cyrille Pitchen <cyrille.pitchen@...e-electrons.com>,
Sagar Dharia <sdharia@...eaurora.org>,
Jens Axboe <axboe@...nel.dk>, guodong.xu@...aro.org,
linux-netdev <netdev@...r.kernel.org>,
Randy Dunlap <rdunlap@...radead.org>,
linux-kernel@...r.kernel.org, Vinod Koul <vkoul@...nel.org>,
linux-crypto@...r.kernel.org,
Philippe Ombredanne <pombredanne@...b.com>,
Sanyog Kale <sanyog.r.kale@...el.com>,
"David S. Miller" <davem@...emloft.net>,
linux-accelerators@...ts.ozlabs.org
Subject: Re: [RFCv3 PATCH 1/6] uacce: Add documents for WarpDrive/uacce
On Mon, 19 Nov 2018, Jerome Glisse wrote:
> > IIRC this is solved in IB by automatically calling
> > madvise(MADV_DONTFORK) before creating the MR.
> >
> > MADV_DONTFORK
> > .. This is useful to prevent copy-on-write semantics from changing the
> > physical location of a page if the parent writes to it after a
> > fork(2) ..
>
> This would work around the issue but this is not transparent ie
> range marked with DONTFORK no longer behave as expected from the
> application point of view.
Why would anyone expect a range registered via MR behave as normal? Device
I/O is going on into that range. Memory is already special.
> Also it relies on userspace doing the right thing (which is not
> something i usualy trust :)).
This has been established practice for 15 or so years in a couple of use
cases. Again user space already has to be doing special things in order to
handle RDMA is that area.
Powered by blists - more mailing lists