On Mon, Mar 5, 2012 at 1:34 PM, Martin Decky <[email protected]> wrote: >> please test/review > > > There is one general issue which needs to be solved before we can integrate > the PCI and USB support for PowerPC into mainline. > > It is imperative that the I/O addresses are not hardwired into the PCI > driver (or the root node driver) because each machine uses different > locations of the PCI buses. Using wrong hardwired addresses can cause the > machine to freeze (there might be some exception which the kernel could > catch to avoid a lockup, but I haven't investigated this yet).
yes, I understand this, I just copied the approach from rootpc, but I guess PCs are safer in this regard, I have no problem with the code sitting in a branch for some time, it would be even better if someone (I don't have any OHCI hw) could test it both on ppc32 and ia32/amd64 (regressions), before mainlining. Another option might be to hide the pci launch code in rootmac driver behind a config option. > > This is the reason for the recent commits in mainline which export the > OpenFirmware device tree into uspace (using sysinfo). In the OFW tree one > can find the PCI buses and their I/O address ranges (see the "ranges" > property in the OFW PCI binding [1]). > > I have already started to implement the decoding of the OFW data in the root > node driver, but it is far from ready yet. is it possible to get the information (in some ugly hackish way) at this stage? i.e. is the information available in the the blob that could be accessed via sysinfo? j.v. > > [1] http://www.openfirmware.org/ofwg/bindings/pci/pci2_1.pdf > > > M.D. > > > _______________________________________________ > HelenOS-devel mailing list > [email protected] > http://lists.modry.cz/cgi-bin/listinfo/helenos-devel _______________________________________________ HelenOS-devel mailing list [email protected] http://lists.modry.cz/cgi-bin/listinfo/helenos-devel
