Re: device detection

[Andrew Clausen]

Glenn McGrath wrote:
> 
> I notice there is FIXME note around pde_devices_probe_all() in devices.
> 
> Currently it just probes a static list of likely target devices, there
> are a couple of problems that i see here.
> 
> 1) Whilst the static list of devices will cover most situations,  it
> wont allways get all devices. Maybe im strange, but i have 6 ide
> channels.

Yep, agreed.  Note: you can do:

        # parted /dev/hdz

Or whatever, if you want.  But, we should try to find all devices
that we can...

> 2) It wont work with devfs, devfs puts device handles in different
> positions.

Yep.
 
> It would be good if parted obtained an acurate list of devices from the
> OS.

Indeed.

> It could be done parsing /proc/devices or /proc/partitions, but this
> method will still fail if the kernel doesnt have /proc support compiled
> in.

/proc is always around, nowadays.  Well, not always, but we can
force everyone to have it!  Not unreasonable IMHO.

(OTOH, FreeBSD is a different story... /proc isn't the same
mess it is on Linux.  /proc is a disgrace, IMHO.  should be /kernel,
or something)

So, how should we parse /proc/partitions?  We need to tell the
partition devices apart from the "real" ones.  Can we simply
strip the numbers off, from the end?

Andrew Clausen