Legacy serial ports detection

[David Zeuthen]

Hi,

Sorry for the lag,

On Sat, 2004-09-04 at 21:20 +0200, Jakub Stachowski wrote:
> Hello, 
> 
> I have made patch for hald that adds support for legacy serial ports (ttySx). 
> For now it has one big problem: it merges also nonexistant devices (for some 
> strange reason sysfs always contains all 'potential' serial ports - in most 
> cases 8 of them - even on my laptop with one real serial port). I can detect 
> them via ioctl but have no idea how to remove 'virtual' ones. I tried 
> overloading class_device_in_gdl and removing there but it hald sometimes 
> crashed then. Important thing is that fake port can be detected only by 
> issuing ioctls over it. I have no more ideas so maybe someone can help me 
> with this.
> 

You should be able to do this in the pre_process method of the
ClassDeviceHandler; just do this

 hal_device_cancel (d);
 hal_device_store_remove (hald_get_tdl (), d);
 return;

if you want to cancel the device (the reason it needs to be done in
pre_process is that we're waiting for e.g udev to name the device).

> As I am completely new to hal some more answers will be appreciated:
> 1) it would be good to detect modem - however it takes quite long, and it 
> would block hald, maybe separate thread?

Uh, that would require locks and everything, probably something we
eventually want to do; we have the same issue with probing storage
devices insofar that it may take several seconds to spin up the drive.

I'm not sure hald should do this automatically when starting up; see
this mail for some thoughts about a Rescan() method:

 http://freedesktop.org/pipermail/hal/2004-September/000871.html

Presumably desktops and operating systems wants to invoke this for the
"Detect Modem" control center thingy. But see below.

> 2) the same goes for detecting printers on LPT (using ieee1284 detection) - 3 
> seconds with my old lexmark 3200. Good think that it won't be frequent - I 
> suppose it could go like this:
>  - try to detect on startup and remember if there was anything
>  - on every tick: try to negotiate 'deviceid' mode - this will fail without      
> device
>  - if last time deviceid mode failed and now succeeded do full detection. This 
> leads to next question:
> What is best method to save detection result between ticks (per port).
>  - if last time deviceid succeeded and now failed remove 'printer' capability
> 
> This scheme fill fail only if someone is capable of replacing driver between 
> two ticks. 
> 
> I suppose similar scheme could be used for detecting modems - using DSR 
> instead of deviceid mode to confirm that there is something to detect.
> 
> 3) Legacy parallel port driver has its own class - 'ppdev' - and printer 
> discovery has to be made on it, not on 'lp' device. I don't know yet how to 
> merge "device" from ppdev class (used for printer discovery) into device from 
> 'lp' class - actual printer. And what to do if lp is merged after ppdev. 
> I have seen patch adding support for legacy lp - is it going to be commited 
> soon?
> 

Very specific questions; some rambling ahead:

Maybe hal shouldn't care too much about legacy ports, perhaps all we
should do is to show that they are available? (e.g. exactly one serial
port on your laptop and zero on my powerbook).

Something more specific like gnome-system-tools, kppp or NetworkManager
should probably use this to help the user detect modems and some of the
existing cups UI software should ask hal for parallel ports to detect
printers? It seems more right to me.. Don't know..

Btw, something like USB modems should probably be autodetected much like
we autodetect USB printers; need to buy an USB modem I guess :-)

Cheers,
David

_______________________________________________
hal mailing list
hal at freedesktop.org
http://freedesktop.org/mailman/listinfo/hal