On Mon, 2012-07-02 at 10:04 +0000, maximilian attems wrote: > On Mon, Jul 02, 2012 at 11:16:49AM +0200, maximilian attems wrote: > > according to media reports several Debian hosted webserver had outage > > on the night to the 1. of July due to the leap second. > > > > according to davej > > http://codemonkey.org.uk/2012/06/15/testing-leap-code/ > > RH uses this fixes: > > fad0c66c4bb836d57a5f125ecd38bed653ca863a > > timekeeping: Fix CLOCK_MONOTONIC inconsistency during leapsecond Fixes a regression introduced by 'ntp: Fix leap-second hrtimer livelock'. > > dd48d708ff3e917f6d6b6c2b696c3f18c019feed > > ntp: Correct TAI offset during leap second Minor. > > 6b43ae8a619d17c4935c3320d2ef9e92bdeed05d > > ntp: Fix leap-second hrtimer livelock The livelock in question involves two threads acquiring two locks in opposite order; prior to 3.4 they were just one lock. But there have been reports of lockup in adjtimex() on earlier kernel versions, which might be a similar bug. > > media reports include > > http://www.heise.de/newsticker/meldung/Schaltsekunde-Linux-kann-einfrieren-1629683.html > > http://www.wired.com/wiredenterprise/2012/07/leap-second-bug-wreaks-havoc-with-java-linux/ > > > > newer patches include these to be watched: > http://thread.gmane.org/gmane.linux.kernel/1321255/ That's the one that caused user-space to effectively busy-wait. Ben. -- Ben Hutchings When in doubt, use brute force. - Ken Thompson
Attachment:
signature.asc
Description: This is a digitally signed message part