Bug#271164: Installation Report amd64(gcc-3.2)

["Harald Dunkel" <harald.dunkel@t-online.de>]

Joey Hess wrote:
> Harald Dunkel wrote:
>
> > - hdparm was not installed. (If you install some non-standard
> >  tools like pppoe, then you should install hdparm, too. I
> >  would guess there are more users with an IDE disk than with
> >  a PPPoE connection.)
>
>
> Hdparm is not necessary if the kernel detects your ide chipset and
> enables dma, which it does for both. Even if hdparm is necessary to turn
> on DMA on your disk, that does not prevent you from being able to
> install. Lack of pppoe support would make it impossible for some users
> to install at all.

Please remember that this was a report about a network
installation. If you could download pppoe via net, then
obviously you were already online.

hdparm is essential on most i386 machines, at least to verify
the settings for your disks. Of course it is not _that_
essential to break anything, if its not available (AFAIK).

> AFAIK we're setting disk labels but have not switched to mounting with
> them, I think it's best to put this change off until after sarge since
> various things can (and in my experience, do) break with disk labels.

Sure.

> > - In d-i I did not disable any of the suggested drivers. Especially
> >  all ide-* drivers were kept enabled. The final system did not use
> >  the expected amd74xx driver for ide, but ide-generic, i.e. dma is
> >  off if you keep the defaults suggested by d-i.
> >
> >  I don't think that this is correct. If I tell d-i to kick out
> >  _all_ suggested ide drivers, then discover loads the expected
> >  amd74xx at boot time, and dma is working correctly.
>
>
> I don't understand this; there is no connection between the code in d-i
> that loads all ide drivers and the code in the initrd of the installed
> system that does the same thing.
This is not about initrd. The IDE drivers are not loaded from the
initrd in this case, because my root disk is managed by sata_sil.
Sorry, probably my report wasn't precise about this.

> I also don't understand why loading adma100, aec62xx, and alim15x3
> before amd74xx would prevent amd74xx from taking over ide for its card,
> or why loading some other drivers and ide-generic later would change
> this.

I am not sure about all the other drivers. Fact is: d-i writes
"ide-generic" into /etc/modules. If I don't manually remove it, then
DMA is turned off for my IDE CDROM. And I can't switch it on using
hdparm. Both IDE drivers (ide-generic and amd74xx) are loaded.

This is the generated /etc/modules for my PC (without the
comment lines):

	ide-cd
	ide-generic
	sbp2
	sd_mod
	sr_mod

AFAICS ide-generic is loaded via /etc/modules very early at
boot time. Later discover finds the amd74xx controller, and
loads the amd74xx driver. But then its too late.


Regards

Harri