Hello! Quoting r. Grant Grundler ([EMAIL PROTECTED]) "Re: [openib-general] Re: mstflint failing on sparc64": > On Thu, Jan 06, 2005 at 03:26:13PM -0800, Roland Dreier wrote: > > > > Generally, do yo happend to know if mmapping /dev/mem > > > > to userspace works on this architecture? > > > > > I can't imagine that it would not. I will see if I can dig info up. > > > The one thing that is weird on sparc64 is that the pci bus is virtually > > > (not physically) addressed. I don't know if you take that into account. > > > > Actually I would suspect that it doesn't. If you look at > > drivers/char/mem.c, > > you can see that there's nothing there to set up the IOMMU, so I don't > > think access to PCI space will work -- a simple remap_pfn_range won't work > > (sparc64 doesn't even define pgprot_noncached). > > IOMMU is only relevant for DMA. I read Tom's reply to mean the host > access PCI bus involves a tranlation as well. e.g a 32-bit bus address > is mapped above 4GB in the host physical address space. > ergo reading the BAR values won't give one the physical address one > needs to access and program device registers. > > grant
So PCI addresses are different from physical addresses, and /dev/mem deals with physical addresses, right? If that's the case, changing mmapped access to pread/pwrite wont help either? So is there a way to get at the pci bus from userspace on that platform? mst _______________________________________________ openib-general mailing list [email protected] http://openib.org/mailman/listinfo/openib-general To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general
