callout api - status report 2

Prashant Vaibhav pvaibhav at freebsd.org
Sun Jun 28 17:25:22 UTC 2009


Hi,
This week I've finished implementation of the modified callout subsystem
which now uses binary heaps. Casual testing shows no signficant issues
anymore. Rescheduling of existing callouts is also now more efficient than
removing/reinserting them, as it now just changes the timeout ticks of the
callout and readjusts the queue as appropriate.
With the basis in place, next week I plan to start implementing the new
callout API. I plan to rewrite a lot of the code from
sys/kern/kern_timeout.c to take advantage of the fact that each operation on
the callout queue leaves it in a consistent state, thus giving us
opportunity to use finer-grained locking.
Perforce is now in sync with my local copy
(//depot/projects/soc2009/calloutapi/) and src/sys can be checked out to
build a test kernel. I plan to resync it every few days (with the goal being
that each version should compile).
I've also started a blog to document the development at:
http://freebsd.mercurysquad.com/
Best,
Prashant


More information about the soc-status mailing list