Messages in this thread | | | Subject | RE: [2.6.25-rc5-mm1][regression] ia64: hackbench doesn't finish >12 hour | Date | Fri, 4 Apr 2008 16:52:29 -0700 | From | "Luck, Tony" <> |
| |
> We can reproduce the soft lockup issue now and root cause the issue as well. > > Since the ptc.g patch uses semaphore ptcg_sem to serialize multiple > ptc.g instructions in ia64_global_tlb_purge(). This requires the code > path should be safe to sleep in down(). But the code path can not > sleep during swap because it holds some spin locks (e.g. anon_vma_lock). > Going to sleep finally causes soft lockup.
I've just pushed a new version of this patch into my test tree (so it will show up in -mm and linux-next when they pull and rebuild). It now uses the Piggin/Wilcox "spinaphore" primitives. It passes all my tests, and some runs of "hackbench" that tried to match the tests that failed before at Fujitsu by forcing moderate amounts of swapping.
-Tony
| |