Virtuous blogs jbarnes' braindump


English (US)   too much travel  -  Categories: News  -  @ 11:24:45 am

Sigh. Oh nevermind, I won’t apologize for my lack of posting. No excuses.

Conferences and planning

So, the past few weeks have been quite busy with travel and hacking. Early in the month I attended the X.Org Developer’s Summit in Edinburgh, Scotland. It turned out to be a very productive conference; we worked out several issues with the DRI2 design, shared a lot of good info, and spent some time planning future releases. Following XDS, I headed up to Portland, OR to meet with some of the Intel driver team to do planning for future development. As usual, there’s a ton of interesting work that needs to be done, so we’ll be spread pretty thin, but it looks like the next year should see some major improvements to all layers of the stack. I’m spending this week in Portland as well, for the Linux Kernel Summit and Linux Plumbers Conference. The kernel summit was fairly good; we even got some hacking & bug fixing done, though we didn’t make much progress (in my opinion) on fixing our procedural issues with profiling & instrumentation. Hopefully someone will step up to take ownership of those areas and get things into shape…


I’ve been careful about pushing PCI fixes for 2.6.27, partly because Linus seems really cranky about post -rc pulls these days and partly because things are actually looking pretty good. The last pull had mostly compile warning fixes (and the current queue includes a couple of those as well) but there weren’t any serious issues that would affect a lot of users to fix once the -rc came out… For 2.6.28 there are a few things I’d like to include, but I have to find time to review and integrate them; we’ll see if I manage to do that (the main ones are VGA arbitration and address space management improvements)


I sent out a 2.5 pre-release for xf86-video-intel awhile back and things have been moving much faster than I’d like since then. Eric fixed up libdrm to separate out the Intel buffer manager API into an Intel specific extension, which is good since we’ll depend on a new libdrm for this release. This ended up destabilizing things a little, but Eric (fortunately given my schedule lately) has been picking up the pieces.

On the DRM front I’ve been working to finalize the GTT page faulting code, which we’ll need for both GEM and kernel mode setting in order to handle fallbacks efficiently in the X server. We may also end up using it to deal with tiling issues. The patch is looking pretty good now (thanks mostly to reviews by Thomas and Nick), I just have one last issue to fix and hopefully we can roll the code into Eric’s GEM branch.

I’m still crossing my fingers that we can get GEM into the drm-next tree in time to get it into 2.6.28. Dave and I harassed some people at KS about the remaining GEM issues, so hopefully we’ll see an ack from Nick on lkml soon, which will allow Dave to pull the bits into his DRM tree. Once that happens, I can diff out the kernel mode setting bits and submit them. Doubtless a few bugs have crept in since I last merged things in the DRM modesetting-gem branch (I think suspend/resume broke along the way somehow) so I’ll have a few things to fix, but other than that things should be in pretty good shape. It would be nice to get this queued for 2.6.28 as well, but that’s fairly ambitious.

In all, I’ll have a ton of catching up to do when I get back home next week. But I’m still holding out hope that I’ll be able to get a little hacking in this week at Plumbers; we’ll see.

Trackback (0)

Trackback address for this post:

This is a captcha-picture. It is used to prevent mass-access by robots.

Please enter the characters from the image above. (case insensitive)


No Trackbacks for this post yet...

This post has 1 feedback awaiting moderation...


No Pingbacks for this post yet...

This post has 1 feedback awaiting moderation...

powered by b2evolution free blog software

Contact the admin - Credits: blog software | blog hosting | Francois