[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-ID: <20200120174933.GB3891@mellanox.com>
Date: Mon, 20 Jan 2020 17:49:39 +0000
From: Jason Gunthorpe <jgg@...lanox.com>
To: Jason Wang <jasowang@...hat.com>
CC: Shahaf Shuler <shahafs@...lanox.com>,
Rob Miller <rob.miller@...adcom.com>,
"Michael S. Tsirkin" <mst@...hat.com>,
"linux-kernel@...r.kernel.org" <linux-kernel@...r.kernel.org>,
"kvm@...r.kernel.org" <kvm@...r.kernel.org>,
"virtualization@...ts.linux-foundation.org"
<virtualization@...ts.linux-foundation.org>,
Netdev <netdev@...r.kernel.org>,
"Bie, Tiwei" <tiwei.bie@...el.com>,
"maxime.coquelin@...hat.com" <maxime.coquelin@...hat.com>,
"Liang, Cunming" <cunming.liang@...el.com>,
"Wang, Zhihong" <zhihong.wang@...el.com>,
"Wang, Xiao W" <xiao.w.wang@...el.com>,
"haotian.wang@...ive.com" <haotian.wang@...ive.com>,
"Zhu, Lingshan" <lingshan.zhu@...el.com>,
"eperezma@...hat.com" <eperezma@...hat.com>,
"lulu@...hat.com" <lulu@...hat.com>,
Parav Pandit <parav@...lanox.com>,
"Tian, Kevin" <kevin.tian@...el.com>,
"stefanha@...hat.com" <stefanha@...hat.com>,
"rdunlap@...radead.org" <rdunlap@...radead.org>,
"hch@...radead.org" <hch@...radead.org>,
Ariel Adam <aadam@...hat.com>,
"jakub.kicinski@...ronome.com" <jakub.kicinski@...ronome.com>,
Jiri Pirko <jiri@...lanox.com>,
"hanand@...inx.com" <hanand@...inx.com>,
"mhabets@...arflare.com" <mhabets@...arflare.com>
Subject: Re: [PATCH 3/5] vDPA: introduce vDPA bus
On Mon, Jan 20, 2020 at 04:43:53PM +0800, Jason Wang wrote:
> This is similar to the design of platform IOMMU part of vhost-vdpa. We
> decide to send diffs to platform IOMMU there. If it's ok to do that in
> driver, we can replace set_map with incremental API like map()/unmap().
>
> Then driver need to maintain rbtree itself.
I think we really need to see two modes, one where there is a fixed
translation without dynamic vIOMMU driven changes and one that
supports vIOMMU.
There are different optimization goals in the drivers for these two
configurations.
> > If the first one, then I think memory hotplug is a heavy flow
> > regardless. Do you think the extra cycles for the tree traverse
> > will be visible in any way?
>
> I think if the driver can pause the DMA during the time for setting up new
> mapping, it should be fine.
This is very tricky for any driver if the mapping change hits the
virtio rings. :(
Even a IOMMU using driver is going to have problems with that..
Jason
Powered by blists - more mailing lists