Messages in this thread | | | Date | Wed, 26 Jun 2013 10:27:40 -0400 | From | Dave Jones <> | Subject | Re: [3.10-rc6] soft lockups from sync() |
| |
On Wed, Jun 26, 2013 at 03:49:15PM +0200, Jan Kara wrote: > On Mon 17-06-13 10:26:11, Dave Jones wrote: > > Haven't seen this happen before.. 3.10-rc6 + a bunch of patches to > > survive other trinity induced bugs + some debug 'helpers' (the > > loadavg-at-oops for eg), but nothing that should be responsible for this. > Hum, we are spinning on inode_sb_list_lock. I'm not aware of any recent > changes in that area. Is this reproducible? Has anything changed with > trinity so that it could start triggering this? Maybe using lockstat we > could see who are the worst offenders using the lock...
The biggest change is that I got two new test machines (that I'm seeing this on). They're both quad core haswell machines, so a lot faster than the previous machines I've been running trinity on.
Dave
| |