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

Re: Bug#416490: acpi-support: IBM Ultrabase X4 DVD/CDOM Undetected On Dock

Hi Henrique,

Henrique de Moraes Holschuh wrote:
On Tue, 03 Apr 2007, Raphael Hertzog wrote:
Debian should be able to handle automatically detecting IDE CDROM's after docking by now, it is 2007 after all!
I would like this too. Unfortunately, I have no clue on how to detect
that we're plugged into a dock. And it's not even clear that acpi-support

Well, since this has "thinkpad" near it, let me make it pretty clear:
whatever you guys do, DO NOT hook into thinkpad-acpi/ibm-acpi bay ejection
and undocking functionality as it stands on anything that will be shipped by
Debian.  I have deprecated it for a long time now, and it *will* be removed
or completely changed (to match generic bay and dock interface) before Lenny
is out.

Thanks for the warning!

Back to the problem at hand:  there are Linux generic ACPI "dock" and "bay"
drivers, they are supposed to notify userspace of any ejections and
reconnections.  Only, they don't do everything that is needed yet, and they
don't handle many of the real ejectable devices in a laptop (e.g. ThinkPad
batteries, floppies, and anything else that is not a SATA/ATA/ATAPI device).

As for automated bus rescanning, sincerely, it is up to the kernel to do it.
Anything else will be just an ugly hack or highly specific to a particular
machine model.  You really need to do ACPI/platform specific node -> kernel
device -> device bus mapping to do it properly, and the kernel is the place
to do it.

Userspace should be getting the usual udev messages for device hot-insertion
and hot-removal, and not need to care about anything else.

If I understand you correctly, your advice would be to tag all of this stuff "wontfix"?

is really the package that should take care of that. After all,
acpi-support is only a set of hacks so that supend/resume works
with as many laptops as possible.

Frankly, a package should deal with anything and everything to do with ACPI
support if it is going to be named "acpi-support".  If it is a
suspend/resume helper, it should have been named differently, IMO.

Raphael's description skipped a few details. First of all, the purpose of the package is not only suspend/resume, it's basically to make as many laptop features work as possible. This includes all of the "special" buttons etc. Yes, this still means acpi-support is a misnomer, but we have an upstream, so package renaming is not a really attractive option. :-)

Anyway, since acpi-support is really a collection of ugly hacks to make various laptops work, and since it is absolutely intended to be a transitional package until better solutions come along for the problems that it solves, there is something to be said for including a dirty hack for common docks as well. OTOH, if this stuff is going to be changed completely before Lenny comes out, it's probably not worth the trouble doing it now.

What I do worry about is that the kernel hotplug events won't materialize in the kernel that will be shipped with Lenny. I don't want to leave these users out in the cold for yet another stable release cycle, so if this stuff isn't fixed in the kernel, I'm tempted to include a hack in acpi-support.


Reply to: