Messages in this thread | | | Subject | Re: [RFC][PATCH 0/3] sched: skip_clock_update madness | From | Mike Galbraith <> | Date | Mon, 05 Jan 2015 13:07:07 +0100 |
| |
On Mon, 2015-01-05 at 12:50 +0100, Peter Zijlstra wrote: > On Mon, Jan 05, 2015 at 11:49:09AM +0100, Mike Galbraith wrote: > > On Mon, 2015-01-05 at 11:18 +0100, Peter Zijlstra wrote: > > > Hi, > > > > > > So the big lockup thread got me looking at the skip_clock_update stuff again > > > and here's what I came up with. > > > > > > It seems to build a kernel without generating lockdep splats. > > > > > > Mike can you see if it cures the wobblies you were seeing? > > > > Nope, those hiccups were production IO beasts from hell staying in the > > kernel for ages at a time. Watchdog being falsely credited with what is > > usually dinky wakeup -> switch delta throttled it effectively forever. > > OK, It's Monday and I'm still trying to bootstrap my post xmas brain. Is > that: Nope - can't check, or: Nope - doesn't fix wobblies ?
That's: nope, the boxen that wobbled to death are not at my disposal.
-Mike
| |