Skip to contentSkip to conversation

Hotplug can configure XFree86

DISCUSSION Stamatis Mitrofanis 4 participants 7 messages Dec 25, 2001 – Dec 30, 2001

lore
Share & tools
lore
#0 lore raw

Hotplug can be made so that it maps PCI device descriptions to records in the XFree86 card databse (/usr/X11R6/lib/X11/Cards) and update the X configuration file accordingly.

This way we can have a nice centralized hardware management system that configures everything. Also, this way it can work flawlessly with the automatic driver download system I wasted (...) so much time on.

One way to do it is to separate the XF86Config file to "hot" parts and a "cold" part. The hot parts are managed by hotplug and are merged with the cold part after every change.

Unfortunately, this won't be very helpful for hotplugging stuff for the X server to use immediately while it is running, since XFree86 just won't reread its configuration file.

What it could do though is set up the X server at boot-time.

I hope you people find it a good idea. I am trying to do an implementation.

_______________________________________________
Linux-hotplug-devel mailing list http://linux-hotplug.sourceforge.net
[email protected]
https://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel

#1 lore raw
Quoted 6 lines

Hotplug can be made so that it maps PCI device descriptions to records in the XFree86 card databse (/usr/X11R6/lib/X11/Cards) and update the X configuration file accordingly.

This way we can have a nice centralized hardware management system that configures everything. ...

I certainly like the notion of the Linux distros having a better integration story in such areas, but I suspect there's a certain matter of installed base to worry about in terms of xf86config, kudzu, YaST, and so forth ...

- dave

_______________________________________________
Linux-hotplug-devel mailing list http://linux-hotplug.sourceforge.net
[email protected]
https://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel

#2 lore raw
In reply to Stamatis Mitrofanis
Quoted 4 lines · David Brownell

I certainly like the notion of the Linux distros having a better integration story in such areas, but I suspect there's a certain matter of installed base to worry about in terms of xf86config, kudzu, YaST, and so forth ...

The installation programs I've seen so far only provided a once-and-for-all static configuration file. I'll be fully satisfied with Linux hardware management when I can freely swap hard drives with any computer and still see the system come up just as it used to.

The ugliest problem to solve is that of the X server not coming up when the graphics adapter has changed. I couldn't find so far a better solution than to integrate some autoconfiguration with the linux-hotplug scripts.

I mentioned that there can be "hot" and "cold" XF86Config parts. The installation programs can manage the "cold" parts which will contain information that linux-hotplug can't (yet) retrieve. Eventually, the "cold" part will become minimal.

_______________________________________________
Linux-hotplug-devel mailing list http://linux-hotplug.sourceforge.net
[email protected]
https://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel

#3 lore raw
In reply to Stamatis Mitrofanis
Quoted 4 lines

I mentioned that there can be "hot" and "cold" XF86Config parts. The installation programs can manage the "cold" parts which will contain information that linux-hotplug can't (yet) retrieve. Eventually, the "cold" part will become minimal.

Well, things that hotplug can't currently guess (as used in an X config)

- monitor type
(which lets you get the available resolutions and depth)
- users's preferred resolution
- users's preferred color depth
- pointer type (for non USB, etc.)

I realize I have a biased opinion, but I don't think that walking the user through a minimal X reconfiguration (as opposed to a completely automated change) is a bad thing on graphics HW change.

Bill

_______________________________________________
Linux-hotplug-devel mailing list http://linux-hotplug.sourceforge.net
[email protected]
https://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel

#4 lore raw
In reply to Stamatis Mitrofanis
Quoted 2 lines

Well, things that hotplug can't currently guess (as used in an X config)

There are lots of things, and some of them _could_ in some cases be autodetected ... or at least old values saved and reused (likely after sanity checking). In particular, don't many of the newer monitors have ways to identify themselves?

Quoted 9 lines

- monitor type
(which lets you get the available resolutions and depth)
- users's preferred resolution
- users's preferred color depth
- pointer type (for non USB, etc.)

I realize I have a biased opinion, but I don't think that walking the user through a minimal X reconfiguration (as opposed to a completely automated change) is a bad thing on graphics HW change.

I don't either, and in fact I'm convinced that there are some configuration actions that deserve user intervention ... even at the level of popping up some dialogue on the user's X11 desktop, and presenting notices (and ideally actions)!

But on the other hand, I also believe that should be avoided whenever possible. So for example in this case, I'd rather see a "looks like you have <this/> new configuration, right?" dialogue rather than doing twenty questions. (Though to be sure, the last few times I had to do the xf86config-ish game the questions were at least ones that mortals might answer!) And sometimes that dialogue would be able to time out without demanding input...

I'm making a fairly general point here, which I hope most folk will agree with: Linux desktop systems should do a lot more "automatic detection" of hardware ("hotplugging") than they've traditionally done, and demand a lot less user input on configuration changes. It's been getting better in the past few years ... and I hope it'll continue !! Likewise, I'll hope that the distros will converge on at least the tools in those areas, if not all the policies they use. That's harder.

- Dave

_______________________________________________
Linux-hotplug-devel mailing list http://linux-hotplug.sourceforge.net
[email protected]
https://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel

#5 lore raw
In reply to Stamatis Mitrofanis
Quoted 11 lines · David Brownell

...

I'm making a fairly general point here, which I hope most folk will agree with: Linux desktop systems should do a lot more "automatic detection" of hardware ("hotplugging") than they've traditionally done, and demand a lot less user input on configuration changes. It's been getting better in the past few years ... and I hope it'll continue !! Likewise, I'll hope that the distros will converge on at least the tools in those areas, if not all the policies they use. That's harder.

Ok... Enough general theory.

The idea was that the hotplug scripts can autoconfigure XFree86. We consider XFree86 to be similar to the kernel (drivers, modules, devices...). The scripts _could_ map actual devices to more abstract XFree86 devices.

I suggest that we do some cherry-picking from open-source installation programs around to find good ways to complete the XFree86 configuration. Then we can include things in hotplug. (right?)

_______________________________________________
Linux-hotplug-devel mailing list http://linux-hotplug.sourceforge.net
[email protected]
https://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel

#6 lore raw
In reply to Stamatis Mitrofanis
Quoted 15 lines · Stamatis Mitrofanis
Quoted 11 lines · David Brownell

...

I'm making a fairly general point here, which I hope most folk will agree with: Linux desktop systems should do a lot more "automatic detection" of hardware ("hotplugging") than they've traditionally done, and demand a lot less user input on configuration changes. It's been getting better in the past few years ... and I hope it'll continue !! Likewise, I'll hope that the distros will converge on at least the tools in those areas, if not all the policies they use. That's harder.

Ok... Enough general theory.

Quoted 8 lines

The idea was that the hotplug scripts can autoconfigure XFree86. We consider XFree86 to be similar to the kernel (drivers, modules, devices...). The scripts _could_ map actual devices to more abstract XFree86 devices.

I suggest that we do some cherry-picking from open-source installation programs around to find good ways to complete the XFree86 configuration. Then we can include things in hotplug. (right?)

Someone has approached [email protected] about this topic, I hope. It would be wonderful to work with the xfree86 folks directly on this.

Miles

_______________________________________________
Linux-hotplug-devel mailing list http://linux-hotplug.sourceforge.net
[email protected]
https://lists.sourceforge.net/lists/listinfo/linux-hotplug-devel