Randy pointed out that I had attached the wrong file to this
message.  This is the correct one.  I don't think it will
apply to the current BK, but just to set the record
straight...

I will have a new one as soon as we iron out the nano_sleep
regs issue.

Have a nice turkey day :)

This patch has no configure options!  

With this version I have changed the signal/ repeating timer
model to avoid timer storms from repeating timers.  The
timer is not inserted back into the timer_list until the
signal is dequeued by the receiving task.  So, once a timer
expires, it will cause NO further system overhead until the
task takes the signal, regardless of the repeat interval. 
The POSIX standard allows this, (actually it seems to
encourage it).  Each timer has an "overrun" counter that
keeps track of the number of times the signal should have
been generated but was not because the prior signal was not
yet picked up.

The timer_gettime() call returns the time as if the timer
were in the timer_list so to the program it looks like it
is, but the system overhead is nil.

The really nice thing about this way of limiting timer
storms is that a really fast real time program can actually
get the high turn over rate while not being impacted by
slower, or blocked programs trying to do the same thing. 
Like wise, if a program loses the cpu, its timers will not
impact other tasks.  Many thanks to Mark Salisbury
<mbs@mc.com> for suggesting this.

The patch stands alone (i.e. does not require any of the
following high-res patches).

This, patch implements the POSIX clocks and timers
functions.  The two standard clocks are
defined(CLOCK_REALTIME & CLOCK_MONOTONIC). 

Nano_sleep() is rolled into clock_nanosleep().  Also a bug
fix in clock_nanosleep().

kernel/timer.c is modified to remove the timer_t typedef
which conflicts with the POSIX standard definition for this
type.  

The patch introduces a new kernel source (posix-timers.c)
which contains most of the code.

This implementation has no limits on the number of
timers in the system or per process or task, thanks to Jim
Houston.

Kernel version 2.5.50


Test programs, man pages and readme files as well as this
patch are available on the sourceforge high-res-timers site: 

http://sourceforge.net/projects/high-res-timers/

Please apply.

-- 
George Anzinger   george@mvista.com
High-res-timers: 
http://sourceforge.net/projects/high-res-timers/
Preemption patch:
http://www.kernel.org/pub/linux/kernel/people/rml