Considering all the talk that has been going on for sometime about kernel
freezes in X etc....I have a problem of mine that I would like input on.
I have ported UTIME (Microsecond Resolution Timers) for linux for 2.1.xx
kernels. What happens is that the kernel freezes within minutes of starting
X. The kernel does not freeze when X is not running (Ive run the kernel
for a couple of days and ive had no problems) The problem seem to be
in the new fast timer code as the same code runs for 2.0.xx kernels
without a hitch.
Here is some information about how I am maintaining usec resolution timers.
In the root timer vector (tv1) I keep the timers sorted based on their
microsecond field. I have a new function (insert_root_timer) which inserts
timers into tv1 in sorted order. run_timer_list then picks out timers from
tv1 based on whether they have expired (ie jiffies + usec)
I cascade timers only when tv1.index==0 and when run timer_list has
wrapped around completely (ie tv1.index has wrapped around to 0 from 255)
Here is my question:
What is so special about X that could cause these crashes??
I would appreciate it a lot if anyone (I know there must be someone in this
list) who knows why this might be happening lets me know.
I can provide more information to anyone that needs it.
I would really like to get this code stabilized so that I can start working
without fear of a crash on my other stuff...
also is there a possibility of getting this code included into the kernel
once it is stabilized??
Thanks in advance
balaji
PS: More information about UTIME can be obtained at
http://hegel.ittc.ukans.edu/projects/utime
-------------------------------------------------------------------------------
Balaji Srinivasan
email: balaji@hegel.ittc.ukans.edu
WWW : http://www.ittc.ukans.edu/~balaji