"Neil A. Carson" wrote:
> 
> The stuff about 'I'll change tsleep to work in an interrupt context'
> scares me somewhat. Doing this requires serious consideration of a
> number of design inssues and IMHO reconsideration of interrupt
> management. Allowing arbitrary drivers to just sleep processes is a very
> bad idea, I think.

The idea here is to make a way to find out we are in interrupt
context and change tsleep() to panic if it's called in interrupt
context. tsleep() can't be made "work in interrupt context",
it needs a process context to sleep on.

Jaromir