Convert the SMC tokenring driver to new probing.

Ugh, two patches got crossed. This is the correct first one.

Originally by Al Viro (NE23-82596)
	* switched 82596 to dynamic allocation
	* 82596: fixed resource leaks on failure exits
Updated to apply agains jgarzik/net-drivers-2.5-exp

>From viro NE22-3c501
	* switched 3c501 to dynamic allocation
	* 3c501: embedded ->priv
	* 3c501: fixed clobbering on autoprobe
	* 3c501: fixed resource leaks on failure exits
Additional:
	* probe correctly when no device present
	* fix loop forever bug in probing
	* free_netdev

Original by Al Viro (NE21-wavelan)
	* switched wavelan to dynamic allocation
	* wavelan: embedded ->priv
	* wavelan: fixed clobbering on autoprobe
	* wavelan: fixed IO before request_region()
	* wavelan: fixed resource leaks on failure exits
	* wavelan: fixed order of freeing bugs
Updated to apply agains jgarzik/net-drivers-2.5-exp

Convert arlan driver to new probing.  This meant a rather large
rework of the probing code for this driver since it did a lot ofnon
standard things.

Originally by Al Viro (NE19-3c507)
	* switched 3c507 to dynamic allocation
	* 3c507: embedded ->priv
	* 3c507: fixed clobbering on autoprobe
	* NB: 3c507.c buggers port 0x100 without claiming it.  Most likely it
	  should be doing request_region() there.
Updated to apply agains jgarzik/net-drivers-2.5-exp

from viro NE18-3c505
	* switched 3c505 to dynamic allocation
	* 3c505: embedded ->priv
	* 3c505: fixed use of uninitialized variable
	* 3c505: fixed resource leaks on failure exits
Additional:
	* add free_netdev

from viro NE17-sk16
	* switched sk_g16 to dynamic allocation
	* sk_g16: embedded ->priv
	* sk_g16: fixed buggy check for signature (|| instead of &&, somebody
	  forgot to replace it when inverting the test).
	* sk_g16: fixed use after kfree()
	* sk_g16: fixed init_etherdev() race
Additional:
	* add free_netdev

from viro NE16-ni5010
	* switched ni5010 to dynamic allocation
	* ni5010: embedded ->priv
	* ni5010: fixed clobbering ->irq
	* ni5010: fixed IO before request_region()
Additional:
	* add free_netdev

>From viro NE15-ni52
	* switched ni52 to dynamic allocation
	* ni52: embedded ->priv
	* ni52: fixed clobbering of everything on autoprobe
Additional:
	* add free_netdev

Convert ni65 driver to new probing; patch sequence goes bottom
up on the probe list.

	* switched ni65 to dynamic allocation
	* ni65: fixed ->irq and ->dma clobbering on autoprobe

Rework de620 driver to new dynamic allocation
Originally by Al Viro.
	* switched de620 to dynamic allocation
	* de620: embedded ->priv
	* de620: fixed IO before request_region()

Updated to ~jgarzik/net-drivers-2.5-exp

New infrastructure to allow probing older builtin drivers (like ISA)
Originally by Al Viro, updated to apply agains jgarzik/net-drivers-2.5-exp

The call to dev_get() in wanrouter_device_new_if is racy and redundant and should
be removed.  The later 'register_netdev()' does the same test internally and will
return the appropriate error if the name already exists.

This patch is against 2.6.0-test6.
Resend of earlier patch because it was ignored, or missed.

Compiles ok (with true .o generated, yeah). Please review.

free_netdev() of devices allocated through use of alloc_netdev().
Though baroque, drivers/net/3c515.c now uses alloc_etherdev().


 drivers/net/3c515.c   |   23 ++++++++++++-----------
 drivers/net/defxx.c   |    2 +-
 drivers/net/dummy.c   |    2 +-
 drivers/net/eql.c     |    2 +-
 drivers/net/ns83820.c |    2 +-
 drivers/net/plip.c    |   14 ++++++++++----
 drivers/net/shaper.c  |   11 ++++++++---
 drivers/net/tun.c     |   18 +++++++++---------
 9 files changed, 43 insertions(+), 31 deletions(-)

Resend of LMC driver patch for 2.6.0-test6
  * do proper probing
  * allocate network device with alloc_netdev
  * use standard pci_id's instead of local defines
  * use standard PCI device interface to find and remove devices.

into home.osdl.org:/home/torvalds/v2.5/linux

Spotted by Matteo Croce 

into home.osdl.org:/home/torvalds/v2.5/linux

From Martin Diehl.

From Martin Diehl.

Also, make sir-dev locking compatible with irport.
From Martin Diehl.

[ATM]: br2684 incorrectly handles frames recvd with FCS (by Alex Zeffertt <ajz@cambridgebroadband.com>)