06.09.2013, в 12:52, Thomas Bogendoerfer <tsbogend@xxxxxxxxxxxxxxxx> написал(а): > On Thu, Sep 05, 2013 at 10:58:25PM +0200, Helge Deller wrote: >> Maybe section 4.5 (Chips et Coherency Issues) is relevant too: ? >> The Rage128 product revealed a weakness in some motherboard chipsets in that there is no mechanism to guarantee > > but radeon is not r128, iirc. There is a discussion on coherency issues with radeon on powerpc platform in linuxppc-embedded mail list. See e.g. here: http://en.it-usenet.org/thread/19030/8457/ I actually tried sync suggestions from there, but they didn't help. > > My current state is: > > - parisc AGP GART code writes IOMMU entries in the wrong byte order and > doesn't add the coherency information SBA code adds I've played with SBA IOMMU code once too. I've added an IOMMU bypass like it done in counterpart ia64 driver, though it didn't help with DRM, plus it broke Fusion MPT SCSI driver, and then i've seen why: /* We are just "encouraging" 32-bit DMA masks here since we can * never allow IOMMU bypass unless we add special support for ZX1. */ In addition, it has nothing to do in AGP mode which is DMA32 limited. > > - our PCI BAR setup doesn't really work very well together with the Radeon > DRM address setup. > DRM will generate addresses, which are even outside > of the connected LBA Aren't they fixing this sort of problems, like here: https://lkml.org/lkml/2010/12/6/516 ? > > I've hacked around both problems, but it doesn't solve the ring test > issue. I even bought an PCI Radeon card to rule out any AGP oddities, > but nothing new came out of the experiments with the PCI card. > > I've started checking drivers/video/aty to see what it does with > acceleration and compare that with radeon DRM. The aty driver uses > an endian config bit DRM doesn't use, but I haven't tested whether > this makes a difference and how it is connected to the overall picture. > > What I'm still wondering is whether radeon DRM really works on 64bit > big endian boxes. Is there any prove, that someone has it running ? > Is it running on any big endian boxes ? I've seen some patches were made for KMS in order to fix support for big-endian machines, however all of them seem to be done for relatively new chips, like r600. All big endian support in r100.c looks like just one buffer swap, AFAIR. > > Thomas. > > -- > Crap can work. Given enough thrust pigs will fly, but it's not necessarily a > good idea. [ RFC1925, 2.3 ] -- To unsubscribe from this list: send the line "unsubscribe linux-parisc" in the body of a message to majordomo@xxxxxxxxxxxxxxx More majordomo info at http://vger.kernel.org/majordomo-info.html
