> Ok, I fixed this and added some speed to add_timer and cascade_timers
> as well. timer.h can now be used as it is in the normal linux
> distribution, the only changes are in kernel/sched.c. The code can now
> be used both in 2.0.28 and 2.1.23 (and most other versions I guess).
>
> Remove the functions add_timer(), del_timer() and run_timer_list() and
> the definitions just before add_timer() from sched.c (they are a
> contiguous block) and replace it by this block:
could 'cascade_timers()' be fixed somehow, to avoid the unpredictable
length cli()-ed loop?
currently cascade_timers() adds unlimited number of timers via
add_timer(). Thats ... ~ 1 millisec cli/sti delay for 1000 timers, ~ 1
second cli/sti delay for 1 million timers. [say when a network link fails
temporarily, all timers count down to zero]
-- mingo