Kurt Van Dijck wrote:
> On Mon, Nov 09, 2009 at 05:09:27PM +0100, Wolfgang Grandegger wrote:
>> Kurt Van Dijck wrote:
>>> On Fri, Nov 06, 2009 at 11:04:27AM +0100, Kurt Van Dijck wrote:
>>>> Thanks to this idea, I added 1 feature to the current softing driver:
>>>> the channel number.
>>>> after this, I think any matching is possible on such devices using udev.
>>> I did some experiments with my udev alternative. seems that the
>>> /sys/class/net/can0/channel file is not present then the daemon wants
>>> it.
>> I still disagree to put files for hardware (non-networking) related
>> properties into
>> /sys/class/net/canX/. What about creating
>> /sys/class/net/device/channelX, or /sys/class/net/device/can0/channnel.
> 
> in fact, /sys/class/net/canX is a symlink to
> /sys/devices/<somewhere>/<device>/net/canX
> 
> on my current system, with softing card, this gives:
> 
> .
> |-- driver -> ../../../../../bus/pcmcia/drivers/softing_cs
> |-- id (new directory, was 'softing')
> |-- net
> |   |-- can0
> |   |   |-- device -> ../../../0.0
> |   |   |-- statistics
> |   |   `-- subsystem -> ../../../../../../../class/net
> |   `-- can1
> |       |-- device -> ../../../0.0
> |       |-- statistics
> |       `-- subsystem -> ../../../../../../../class/net
> `-- subsystem -> ../../../../../bus/pcmcia
> 
> 1) creating /sys/devices/<somewhere>/<device>/channelX feels like
> re-doing the class device work (net/canX).
> 
> 2) to solve the 'multiple bus on a single card' issue, even creating
> /sys/class/net/canX/device/channelX does not solve the problem.
> udev scripts would only know to start from /sys/class/net/canX.
> Once you follow the device/ symlink, you loose the channel information.
> At least, you need the channel number, or another symlink in the
> /sys/class/net/canX directory.
> 
> 3) my current can devices have properties like address, addr_len,
> features, which are very hardware related though.
> 
> I would prefer dividing properties into device specific & network
> specific.
> 
> Having said this, would consulting netdev mailing list be an option for
> this?

Yes, that's a good idea even if I'm not sure that we will get a usable
feedback. Please try.

Wolfgang.
_______________________________________________
Socketcan-core mailing list
[email protected]
https://lists.berlios.de/mailman/listinfo/socketcan-core

Reply via email to