Messages in this thread | | | Date | Tue, 12 May 2015 18:07:37 -0700 | From | josh@joshtrip ... | Subject | Re: [PATCH tip/core/rcu 0/13] Torture-test updates for 4.2 |
| |
On Tue, May 12, 2015 at 03:57:53PM -0700, Paul E. McKenney wrote: > Hello! > > This series contains torture-test updates: > > 1. Use correct locking primitive, thus avoiding deadlock, courtesy > of Alexey Kodanev. > > 2. Exchange TREE03 and TREE04 geometries in order to improve bug-location > rates. > > 3. Exchange TREE03 and TREE08 NR_CPUS, speed up CPU hotplug to further > improve bug-location rates. > > 4. Allow negative values of nreaders to in order to oversubscribe > the CPUs, again to improve bug-location rates. > > 5. Change longdelay_us to longdelay_ms in order to better reflect > reality. (Not that gcc cares, but people reading the code just > might.) > > 6. Replace open-coded memory barriers with the shiny new > smp_store_release() and smp_load_acquire() primitives. > > 7. Test SRCU cleanup code path in order to improve test coverage. > > 8. Avoid explicitly setting CONFIG_TASKS_RCU, given that it is > now set implicitly based on other configuration options. > > 9. Update configuration fragments from CONFIG_RCU_FANOUT_EXACT to > the new rcutree.rcu_fanout_exact= boot parameter. > > 10. Make rcutorture scripts force RCU_EXPERT so that they can still > build the needed kernel configurations. > > 11. Update TREE_RCU-kconfig.txt to describe new kernel-parameter > setup. > > 12. Make torture scripts display "make oldconfig" errors > > 13. Allow repetition factors in Kconfig-fragment lists. Because > typing 48 repetitions of "TINY02" is getting old.
I replied to patches 4 and 11 with feedback. For the rest:
Reviewed-by: Josh Triplett <josh@joshtriplett.org>
| |