|
From: Eric W B. <ebi...@ln...> - 2002-09-28 06:20:08
|
ke...@us... (Ken Yap) writes: > >In 5.1 I have already moved the pci_ids into the driver files so > >auto-generating config.c would be moving in the opposite direction, > >and it would would make it very hard to have a multiple driver etherboot. > > I know, but I'm not autogenerating config.c, just include files called > $family_ids.c containing the table so you can just as easily include the > file in $family.c and ignore config.c. I agree that it makes more sense > to have the table in $family.c, all you need is a Perl fragment to scan > for a known pattern indicating a table and to turn that into NIC lines. I will have to see what can be done. For now I am knocking off bugs as I come to them. And manually maintaining NIC for a little while longer isn't a huge problem. > BTW are you keeping track of which drivers are working in 5.1? All of the drivers should still be working minimally in 5.1. Though I may have fat fingered the prism driver. I know I touched it just a little bit, adding/updating the virt_to_bus/bus_to_virt code. The drivers that work with multicast are marked as such. I have don't anything as far as testing which drivers work with relocation. I suppose I should go through the drivers and add: #ifdef RELOCATION #warning this driver has not yet been tested with relocation enabled. #endif And then we can remove those lines as the drivers are tested. And I should probably update LOG as well to note all of the other things that have happened in the development branch. At the moment, I am going home for bed. The multicast code it has been working solidly on a 960 node cluster, with no real complains. On Wednesday I am scheduled to go through and do some reboot testing. To test BIOS stability. But if I have missed anything in the multicast slam protocol it will show up there as well. Eric |
