Re: USB device allocation

Khimenko Victor (khim@sch57.msk.ru)
Fri, 8 Oct 1999 18:24:03 +0400 (MSD)


In <19991008020116.C14664@lamia.loth.demon.co.uk> Steve Dodd (dirk@loth.demon.co.uk) wrote:
> On Fri, Oct 08, 1999 at 09:38:00AM +1000, Nathan Hand wrote:

>> The /dev directory can still be "dynamic", the difference is that it's an
>> ext2 filesystem being updated from userspace, rather than a virtual devfs
>> filesystem being updated from kernelspace.
>>
>> Realistically it only loses two features that people might covet (special
>> behaviour on open() of non-existent nodes, ability to use a non-UNIX like
>> filesystem for your /). The majority of people lose nothing.

> Erm, doesn't the first one break module autoloading? I would have thought
> quite a lot of people used that.

> AFAICS what's needed is to separate device detection from device handling -- I
> don't know how much of this is done already. When devices are detected
> (normally boot time except for hotplug buses), poke the user space daemon
> to create the device special file, but don't load the driver. When the device
> file's accessed, the normal kerneld / kmod mechanism kicks in to load the
> driver.

Unfortunatelly it's VERY common scenario here: just power on SCSI scaner after
starting up comp. Even if it's not physical connection it's still new device
for system. And WITHOUT any notification, btw !

-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/