maximilian attems wrote: > initrd-tools loads _all_ ide driver and lets them fight out: > the ones which don't unload stay. initrd-tools loads ide-generic last, AFAIK this makes it only take over and disable DMA if no other module supports the hardware. > the bad side effect of a winning ide-generic is !dma on that box > http://lists.debian.org/debian-kernel/2005/06/msg00100.html That's not how I read that message, in particular it contains the text "If the driver was _not_ compiled as module then the kernel code would determine that DMA was not a supported feature and prevent it from being ever used." (empasis mine) > regarding dma the newer initramfs generating tools: > yaird has troubles with the load ordering. > initramfs-tools did hardcode ide-generic until 0.40 > and had dma troubles too. (see #336103) > > newer udev doesn't bring up ide-generic. > initramfs-tools fails to boot on those without ide-generic > explicitly added to /etc/mkinitramfs/modules. > (wide range of hardware like thinkpad r40e) Yes, the need to hardcode ide-generic to load after the other modules introduces complexity. d-i had to deal with this too (see archives from 2003). > so currently i'd vote for removing that patch. > happy to hear your opinion? :) Aside from reducing the chances that we will ever support 2.6 boot floppies (since building the modules in makes the kernel image even less likely to fit, to the extent that even a bare zImage might not fit on a floppy anymore), dropping them from 2.6 seems like a good idea. If the patch were also dropped from 2.4 I think it would cause the current i386 boot floppy to be overfull by approximately 200k, although there are other changes that could be made to the kernel to partially offset this, such as moving the one kernel symbol used by usb-storage that unnecessarily pulls in the whole ide-core module. -- see shy jo
Attachment:
signature.asc
Description: Digital signature