So I merged the rewrite of the radeon/r200/r300 to the mesa master today.
On a non-KMS DRI1 system we are hoping these patches are mostly regression free in terms of lockups and possibly fix a fair few lockups thanks to Maciej and Jerome's work.
Of course on a KMS/DRI2 system, this work enables GL on r100-r500 families of radeon cards and we hope to add more features to the ones we gain. We should now at least have FBOs where we never did before (in DRI2/KMS mode).
R600 work is ongoing in a branch still so this announcement has no effect on r600 or above.
I've also sent Linus a drm pull request with all of Intels drm-intel-next branch + all the patches I had sitting in my queue.
Early next week I'll send him the KMS patches for radeon, which will be in the radeon driver but the enable switch will be hidden in staging, while we stabilise it, since its such a huge amount of code. So don't expect miracles out of it anytime soon.
On a PCI ID table comparison Intel KMS have 28 IDs to support, radeon KMS has ~350, granted this code has no memory manager for r600 and up yet (kms code sohuld all be there), so its probably close to 200 GPU variants (+/- 50 for ids not seen in the wild maybe).