PM> That's the bad solution. Right solution is to promote threads that
PM> are in kernel mode to non-SCHED_IDLE priority. It should be well
PM> doable, and IMO is required for patch to be usable.
As it is, the kernel locks often become a bottleneck on systems with
lots of penguins. Look at this scenario:
1) A process currently sleeping in kernel mode is ready to go forward
2) At least one processor is busy running a SCHED_OTHER process
Would it benefit throughput if it was guaranteed that the sleeping
process gets to go forward at the expense of the already running
process? It would mean that processes get out of kernel mode quicker.
It would certainly be unfair, so there would have to be special
protections in place to ensure that the process that benefits gets to
pay later.
Benny
-
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@vger.rutgers.edu
Please read the FAQ at http://www.tux.org/lkml/