Loading ide modules from /etc/modules causes DMA to be disabled

[Matt Zimmerman]

I have a system whose root is on a SCSI disk, but which also has an IDE disk.
As a fix for Bug #8156, debian-installer will write ide-disk to /etc/modules so
that these disks are recognised and available in the installed system.

However, the situation which resulted (at least for me) was that DMA was
disabled on the IDE disk. The sequence of events was:

ide-disk loaded from /etc/modules
ide-generic loaded from /etc/modules
via82cxxx loaded by hotplug

The resulting messages from via82cxx were:

Sep 23 00:11:22 mizar kernel: VP_IDE: IDE controller at PCI slot 0000:00:07.1
Sep 23 00:11:22 mizar kernel: VP_IDE: chipset revision 6
Sep 23 00:11:22 mizar kernel: VP_IDE: not 100%% native mode: will probe irqs later
Sep 23 00:11:22 mizar kernel: VP_IDE: VIA vt82c686b (rev 40) IDE UDMA100
controller on pci0000:00:07.1
Sep 23 00:11:22 mizar kernel: VP_IDE: port 0x01f0 already claimed by ide0
Sep 23 00:11:22 mizar kernel: VP_IDE: neither IDE port enabled (BIOS)

Is the bug that ide-generic is loaded first? Should ide-generic not be added to
/etc/modules? hotplug loads the correct module later, and if I add via82cxxx to
the top of /etc/modules, ide-generic is unused and can be removed.

[Herbert Xu (herbert-gondor)]

(In reply to comment #0)
>
> However, the situation which resulted (at least for me) was that DMA was
> disabled on the IDE disk. The sequence of events was:
>
> ide-disk loaded from /etc/modules
> ide-generic loaded from /etc/modules
> via82cxxx loaded by hotplug

For onboard chips the chipset driver must be loaded before ide-generic.

[Matt Zimmerman (mdz)]

(In reply to comment #1)
> For onboard chips the chipset driver must be loaded before ide-generic.

I think it should only affect systems with both SCSI and IDE disks, so I've
downgraded the severity. I suppose this could be fixed in the long term with
hotplug-in-initrd, but it would be nice if we could work around it somehow for
Warty. One idea would be to delay loading of ide-generic somehow until after
hotplug in the boot process

[Herbert Xu (herbert-gondor)]

(In reply to comment #2)
> Warty. One idea would be to delay loading of ide-generic somehow until after
> hotplug in the boot process

Yes that should work. Or you could not load ide-generic/ide-disk at all but
always create the device in udev to let the old autoload mechanism handle it.
You'll need to fix the alias in module-init-tools to point to ide-generic
instead of ide-probe-mod for this to work.

[Matt Zimmerman (mdz)]

This is really the same thing as #1763; if everything were autoloaded correctly,
this would be a non-issue.

This bug has been marked as a duplicate of bug 8509.