Re: request_firmware() hotplug interface, third round.

[Jean Tourrilhes <jt@bougret.hpl.hp.com>]

On Fri, May 16, 2003 at 05:53:17PM +0200, Oliver Neukum wrote:
> >
> >  Hotpluggability is not required, it is the same for any module, which
> >  gets loaded while the system is running. Drivers don't even need to be
> >  aware of hotplug.
> 
> In that case they can contain the firmware and mark it __init.
> They need no RAM. They can even mark the code needed to put
> firmware into the device as __init.

	I think you missed a lot of the previous discussion. Many high
level kernel developpers (Jeff, Alan, Greg) have said 'niet' to binary
firmware linked with the kernel (unless there is source code
available). Please assume that all drivers currently including
firmware blobs in the kernel will need fixing and therefore should not
be taken as example.
	The threads :
		http://marc.theaimsgroup.com/?t=105222131600002&r=1&w=2
		http://marc.theaimsgroup.com/?t=105045487300002&r=1&w=2
		http://marc.theaimsgroup.com/?t=105181861200001&r=1&w=2
		http://marc.theaimsgroup.com/?t=105295022700002&r=1&w=2
		http://marc.theaimsgroup.com/?t=105302951900003&r=1&w=2
	Yes, that's a lot to read, but this will explain the full
complexity of the situation.

> An awful lot of overhead.

	It's not like we have a choice on this issue. The embedded
people are already grumbling.

> 	Regards
> 		Oliver

	Jean