Date: Sat, 19 Dec 2009 21:07:38 -0500 From: Steve Polyack <korvus@comcast.net> To: Robert Noland <rnoland@FreeBSD.org> Cc: freebsd-x11@FreeBSD.org, 'John Baldwin' <jhb@FreeBSD.org>, 'Andriy Gapon' <avg@icyb.net.ua> Subject: Re: PCI Radeon 9250 - DRI/DRM in 8.0-RELEASE Message-ID: <4B2D86EA.5010805@comcast.net> In-Reply-To: <1261246173.2315.20.camel@balrog.2hip.net> References: <200912111614.47417.greg@rowes.org> <4B22C93D.9000208@icyb.net.ua> <4B2317CF.5080900@comcast.net> <E3A590744E63446287A7C170D93388FD@GregPC> <1260636181.2281.35.camel@balrog.2hip.net> <7EE7B533BABE48F7876B0560C84A8238@GregPC> <1261246173.2315.20.camel@balrog.2hip.net>
next in thread | previous in thread | raw e-mail | index | archive | help
On 12/19/2009 1:09 PM, Robert Noland wrote: > On Sat, 2009-12-12 at 17:36 -0500, Greg Rowe wrote: > >> The system is still running the pagezero patch if that makes any difference? >> > Ok, you can get rid of that. > > I wasn't able to fix this quite like I wanted to, but this should work. > The fix that I was intending to do, ended up meaning that I had to work > on our mmap implementation, which I did, but won't be committed until it > is reviewed. It also required a handful of changes to userland (libdrm, > and the DDX driver) and broke the KBI so rebuilding even more stuff was > needed. The linux guys would not have been happy with me and I would > have had a fight on my hands. > > This patch implements what I wanted to do on 64 bit platforms. On 32 > bit platforms (i386) it comes with some consequences... On amd64, I've > adapted the map handle to have an offset of 0 and a globally unique map > number starting at bit 40. This means that there are 24 bits for maps > and each map can describe 1TB of memory. On i386, I've done the same > thing except I am only stealing the top 4 bits for the map id, so there > can only be 15 maps each up to 256MB. This should be ok since normally > there are only 6 or 7 maps. > > I have run tested this on r600 amd64 so far. I can test r3/5/6/700, > Intel and nouveau, but haven't done so yet. This should be a kernel > only fix, meaning you should not need to patch or rebuild any of your > userland apps. The patch is applicable to -CURRENT, and should work on > 8-STABLE and possibly 7-STABLE. > > http://people.freebsd.org/~rnoland/drm_mmap_fix.patch > > robert. > I'm syncing up to 8-STABLE now and doing the builds, but I won't have any results back to you until Monday. I'll let you know if it fixes my issue. Thanks again, Steve
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?4B2D86EA.5010805>