From mboxrd@z Thu Jan 1 00:00:00 1970 From: Ian Campbell Subject: Re: [PATCH V2 3/4] xen/arm: Add support for Broadcom 7445D0 A15 platform. Date: Fri, 3 Oct 2014 16:22:59 +0100 Message-ID: <1412349779.12695.29.camel@citrix.com> References: <542C1D80.6070804@linaro.org> <1412288214-19224-1-git-send-email-jfraser@broadcom.com> <1412288214-19224-4-git-send-email-jfraser@broadcom.com> <542E7E24.60500@linaro.org> <3D592480DD42CA41814EC9221D46F45931DDA2EC@SJEXCHMB12.corp.ad.broadcom.com> Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: <3D592480DD42CA41814EC9221D46F45931DDA2EC@SJEXCHMB12.corp.ad.broadcom.com> List-Unsubscribe: , List-Post: List-Help: List-Subscribe: , Sender: xen-devel-bounces@lists.xen.org Errors-To: xen-devel-bounces@lists.xen.org To: Jonathan Fraser Cc: Julien Grall , "tim@xen.org" , "stefano.stabellini@eu.citrix.com" , "xen-devel@lists.xen.org" List-Id: xen-devel@lists.xenproject.org On Fri, 2014-10-03 at 15:15 +0000, Jonathan Fraser wrote: > On Fri, 2014-10-03 at 11:44 +0100, Julien Grall wrote: > > Hello Jon, > > > > On 02/10/2014 23:16, Jon Fraser wrote: > > > This code supports the Broadcom 7445D0 32-bit A15 based platform. > > > > While this patch has been acked and applied. Did you check that the (by > > default mapped 0xb0000000-0xb0020000) doesn't overlap to the RAM or > > another device on your platform? > > > > Regards, > > > > I had a question about that. This is a guest physical address, not the > machine physical address? It's guest physical but dom0 is mapped 1:1 so that is somewhat the same as machine physical in this instance. > What are the restrictions? It just needs to be a hole in the address space which isn't used for anything else, so not RAM and not MMIO. Since dom0 is 1:1 mapped for MMIO this means a hole in the machine address map, or as a last resort an address used by something which isn't given to dom0. In 4.6 we are going to sort out a way of finding these addresses automatically, so the suckitude of having to figure one out will go away... > On our current platforms, memory always starts at 0 and we currently > support between 1 and 6 GB. There's always a 1GB hole at 0xc0000000. I suppose this is an MMIO hole? Any free space in thatregion would do. > However in the future, memory will start at 0x40000000 and extend > contiguously past 32-bit space for up to 4GB of memory, then segmented > in 36-bit address space. In this case I guess the MMIO has moved to 0...0x40000000? Might need to move the gnttab then too, or maybe we'll have fixed this in the core by then (fingers crossed!) > I haven't tried a guest with much more than 1GB of memory, but will > try that today. > > Thanks, > Jon > >