[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Improving hardware support in Debian stable



This mail is a follow-up to yesterday's meeting at DebConf:
https://penta.debconf.org/penta/schedule/dc13/event/1013.en.html
(recordings not yet available, but should be soon)

I don't think anyone took notes at the time, but I think there was
general consensus that:

1. compat-drivers should be packaged separately from linux, and should
build a separate binary package per subsystem (and per flavour).  This
should remove the risk of regressions for hardware that's already
supported.

2. The installer should install whichever of these binary packages are
needed (presumably with the aid of discover).

3. Mesa drivers for new chips should be packaged with different names
and only bound to the new PCI IDs.  Mesa can then automatically load
whichever driver is appropriate.

4. The new/updated drivers potentially get updated at every point
release.  We won't attempt to support multiple new versions of a driver
per suite.  There is a risk of regression for those using them, but this
is no worse than the current situation where users must update the
entire kernel or X server from another suite which is updated even more
frequently.

I don't think X video drivers and libdrm modules were explicitly
addressed, but presumably they can be handled similarly to Mesa drivers?

Hardware support in the installer itself wasn't addressed within the
meeting proper, but after later discussion it was obvious that
compat-drivers must build udebs to be included in installer images.

Now there is a potential problem with the installer, in that the updated
kernel modules would always be 'installed' in the initrd, whereas on the
installed system they would only be installed if necessary.  So if a
backported driver has a regression in support for older hardware, it
would not affect the installed sytem but *would* affect the installer.

I'm not sure how to get around that.  Possibly the new modules should be
renamed and have their PCI IDs restricted.  But this is not ideal as it
will force people to configure module options differently, and it will
require new logic in initramfs-tools to find the updated drivers.  I'll
keep thinking about how to solve this, but would appreciate suggestions
from others...

Ben.

-- 
Ben Hutchings
Experience is what causes a person to make new mistakes instead of old ones.

Attachment: signature.asc
Description: This is a digitally signed message part


Reply to: