It's amazing what you spot when you Lindent things.

Modern binutils spits out a warning, fixing this broke older binutils.
I guess we can do this in 2.5 ?

From 2.4

Originally from Matt Domsch.
Adds a nowayout module option to override CONFIG_WATCHDOG_NOWAYOUT
From 2.4

From 2.4

Recommended version bump.

Original from William Lee Irwin III <wli@holomorphy.com>

s/more then/more than/

Originally from Matt Domsch.
Adds a nowayout module option to override CONFIG_WATCHDOG_NOWAYOUT
From 2.4

Originally from Matt Domsch.
Adds a nowayout module option to override CONFIG_WATCHDOG_NOWAYOUT
From 2.4

Originally from Matt Domsch.
Adds a nowayout module option to override CONFIG_WATCHDOG_NOWAYOUT
From 2.4

Originally from Gerd...

I've just noticed that a hole left in the recent changes which should
allow the usb v4l drivers to unregister with open file handles.  The
drivers itself handle it just fine, but video_generic_ioctl() will barf
when called on unregistered devices.  Oops.

One way to fix this is to expect drivers call the helper function and
pass a pointer for the function doing the actual work, i.e. handle it
this way:

driver_ioctl(inode,file,cmd,userptr)
	-> video_usercopy(inode,file,cmd,userptr,func)
		copy_from_user(...)
		-> func(inode,file,cmd,kernelptr);
		copy_to_user(...)

Patch against 2.5.7-pre2 below. It updates videodev.[ch] and adapts
usbvideo.c to show how the driver changes will look like.

Note that this change makes the usercopy helper function a very generic
one, it probably could be used for other drivers to (as long as the API
has sane magic numbers based on _IO*(...) defines) as there is no
video4linux-related stuff in there any more.  So we might think of
renaming it an moving it to some more central place (fs/ioctl.c maybe).

Originally from Matt Domsch.
Adds a nowayout module option to override CONFIG_WATCHDOG_NOWAYOUT
From 2.4

Mostly from 2.4. Adds some new ones, and removes some dupes.

A few follow ups to the cleanup done circa 2.5.4

 From Peter Osterlund <petero2@telia.com>

I can't get UDF write support to work in kernel 2.5.7 or 2.5.7-pre2.
The problem is that linux/config.h is not included, so CONFIG_UDF_RW
is undefined and the driver is compiled without write support. This
patch fixes my problem:

Mostly harmless but 1-2 break compilation.

From: Michael Haardt <michael@moria.de> for 2.4, munged by me.

the MAD16 driver is able to enable/disable its gameport, but it does
not register it in the input subsystem.  The appended patch against
2.4.19-pre4 fixes that and also allows to enable to gameport via kernel
command line, not only when loaded as module.