Trying a network install of 2007-01-27 rawhide, I get no hard drives found. I have an nVidia based mboard with an MCP2A IDE controller (PCI ID 10DE:0085). This is apparently supposed to be driven by pata_amd (it is listed in modules.alias with that PCI ID), but it didn't get loaded. If I load it manually, my hard drive and DVD-ROM drive are found. It did load sata_nv, but I don't currently have any SATA drives attached.
Hrmm... are the ids with sata_nv the same? If so, we may need to extend the hackery we previously did for ahci/ata_piix
sata_nv has: alias: pci:v000010DEd*sv*sd*bc01sc01i* alias: pci:v000010DEd*sv*sd*bc01sc04i* aliases. I suppose that's mucking things up.
Yep, sata_nv.c has: static const struct pci_device_id nv_pci_tbl[] = { ... { PCI_VENDOR_ID_NVIDIA, PCI_ANY_ID, PCI_ANY_ID, PCI_ANY_ID, PCI_CLASS_STORAGE_IDE<<8, 0xffff00, GENERIC }, Since it doesn't work with all nVidia IDE devices, that seems broken to me. Is there not a difference between IDE and SATA that should be here?
CC'ing relevant kernel people.
Amusingly (or perhaps not so much), NVIDIA added that generic entry in order to handle all future devices they would produce. The presumption being that [anaconda | hwdata | whomever] should first match the more specific entries such as those found in pata_amd, before a generic fallback entry.
*** Bug 226912 has been marked as a duplicate of this bug. ***
*** Bug 227441 has been marked as a duplicate of this bug. ***
*** Bug 228770 has been marked as a duplicate of this bug. ***
*** Bug 231557 has been marked as a duplicate of this bug. ***
*** Bug 233723 has been marked as a duplicate of this bug. ***
Are steps being made towards fixing this? Will it be fixed within anaconda or what? I would really say sata_nv is broke, as it claims PCI devices it cannot handle (which is kind of the case with ahci as well since it can't handle some devices in some modes but claims them anyway). The "quick fix" for sata_nv vs. pata_amd may just be to have anaconda and such choose the more specific match when looking at PCI IDs.
After finally tracking down a machine, added code in kudzu so that the more specific match gets picked rather than just the first.
*** Bug 234583 has been marked as a duplicate of this bug. ***
*** Bug 234590 has been marked as a duplicate of this bug. ***
*** Bug 234832 has been marked as a duplicate of this bug. ***
*** Bug 234660 has been marked as a duplicate of this bug. ***
*** Bug 235002 has been marked as a duplicate of this bug. ***
*** Bug 235407 has been marked as a duplicate of this bug. ***
Created attachment 152019 [details] crash dump from install
*** Bug 236519 has been marked as a duplicate of this bug. ***
Bug is no longer present in test 4
I am no longer seeing the bug I reported either. I think the assessment and fix were accurate. Thank you.