diff options
author | Patrick Williams <iawillia@us.ibm.com> | 2014-07-17 14:01:40 -0500 |
---|---|---|
committer | A. Patrick Williams III <iawillia@us.ibm.com> | 2014-07-17 16:03:02 -0500 |
commit | cbcc22add05fb4a43faed34edd2b1ffa039dfc23 (patch) | |
tree | 74b26fb64fc221d58c66e94da3f090fa9e10e62d /src/usr/intr | |
parent | 89bd6adf7689cbddad15cd9a9e55d4b98b1df183 (diff) | |
download | talos-hostboot-cbcc22add05fb4a43faed34edd2b1ffa039dfc23.tar.gz talos-hostboot-cbcc22add05fb4a43faed34edd2b1ffa039dfc23.zip |
Restrict timebase sync to once per core.
As part of Id3a3bc0b7367e61f2725af17975fe3ba068f69a9, I fixed the
deferred work queue to not leak work objects if there are multiple
objects. When cores wake up, each thread inserts a work object to
synchronize its timebase. Now that they are not leaking, we are
running this 8 times, which is causing enough clock drift that we
are getting passed the timeout for core wakeups.
Modify deferred work queue to allow us to skip performing work
if there is already an outstanding deferred work object. This
will return us to running the timebase sync just once.
Change-Id: Iccffeb9d0578dcd08d41d41ca6af1b82388e7e34
RTC: 111512
Backport: release-fips811
Reviewed-on: http://gfw160.aus.stglabs.ibm.com:8080/gerrit/12281
Tested-by: Jenkins Server
Reviewed-by: Brian Silver <bsilver@us.ibm.com>
Reviewed-by: A. Patrick Williams III <iawillia@us.ibm.com>
Diffstat (limited to 'src/usr/intr')
-rw-r--r-- | src/usr/intr/intrrp.H | 3 |
1 files changed, 1 insertions, 2 deletions
diff --git a/src/usr/intr/intrrp.H b/src/usr/intr/intrrp.H index 8a7360828..314ae5217 100644 --- a/src/usr/intr/intrrp.H +++ b/src/usr/intr/intrrp.H @@ -301,8 +301,7 @@ namespace INTR enum { - // TODO: RTC 111512; change back to 1s - CPU_WAKEUP_SECONDS = 2, + CPU_WAKEUP_SECONDS = 1, CPU_WAKEUP_INTERVAL_COUNT = 10, CPU_WAKEUP_INTERVAL_NS = (NS_PER_SEC * CPU_WAKEUP_SECONDS) / CPU_WAKEUP_INTERVAL_COUNT, |