Alexander Graf <ag...@suse.de> writes: > On 05.07.2012, at 18:20, Markus Armbruster wrote: [...] >> When it comes to block device models, IDE is *always* the troublemaker. >> And a big reason for that is the messy data structures that are >> impractical to clean up while we still support non-qdevified IDE. Which >> we do almost three years after IDE qdevification. >> >> The laggards are: >> >> * mac99, g3beige >> Alexander Graf <ag...@suse.de> >> >> * spitz, borzoi, terrier >> Andrzej Zaborowski <balr...@gmail.com> >> >> * tosa >> unmaintained >> >> * r2d >> Magnus Damm <magnus.d...@gmail.com> >> >> Any plans to drag these boards into the current decade already? > > Got a pointer to a board that does it right?
Any board that uses pci_cmd646_ide_init(), pci_piix[34]_ide_init(), or vt82c686b_ide_init(). These create a qdev providing an IDE bus. The non-qdevified IDE controllers use ide_init2_with_non_qdev_drives() instead. Funny: g3beige has two IDE controllers, first one isn't qdevified (created with pmac_ide_init()), second one is (created with pci_cmd646_ide_init()). Feel free to ask for more help!
