Messages in this thread | | | Date | Mon, 11 Jun 2012 11:01:12 -0700 | From | "Paul E. McKenney" <> | Subject | Re: rcu: endless stalls |
| |
On Mon, Jun 11, 2012 at 07:20:14PM +0200, Mike Galbraith wrote: > On Mon, 2012-06-11 at 16:22 +0200, Mike Galbraith wrote: > > On Mon, 2012-06-11 at 06:39 -0700, Paul E. McKenney wrote: > > > On Mon, Jun 11, 2012 at 12:06:16PM +0200, Mike Galbraith wrote: > > > > Greetings, > > > > > > > > I received a report of a 48 core UV box hitting a gripe, taking longer > > > > than timeout to emit same, so box griped endlessly, forcing reboot. > > > > > > So it took minutes to print an RCU CPU stall warning? On only 48 CPUs? > > > > > > If so, yow!!! > > > > The report was a tad fuzzy on details, but serial console can take a > > while at low baud rate, and the box was apparently perma-stalled. > > Gee, no wonder it took a while to emit the gripe, there were a few more > than 48 CPUs actually.. 4048 more to be precise. > > 4096 tasks on 4096 CPUs all hammering wait_lock at once r!pretty.
That would be bad...
BTW, one other thing on the patch -- it needs to use a separate flag, otherwise there are races with sysfs and panic() updates to the flag.
> 2aa15890 - mm: prevent concurrent unmap_mapping_range() on the same inode
I confess, you lost me on this one. You believe that this commit is the cause of the RCU CPU stall warnings?
Thanx, Paul
| |