RFC Refcount based Cpu Hotplug. V2
Get series with b4
b4 am [email protected]Versions newest first
- v1 RFC Latest5 patches 10 repliesIn review
Changes since v1 from the cover letter
- !CONFIG_HOTPLUG_CPU part is now handled correctly, thanks to the patch from Paul Jackson.
- The cpu_hotplug_begin() uses a waitqueue instead of a completion struct where a writer can wait while there are active readers in the system.
- Provided a new API's cpu_maps_update_begin(), cpu_maps_update_done() for serializing the updates to cpu_present_map and cpu_online_map. Thus threads which update the cpu_present_map should now call cpu_maps_update_begin instead of lock_cpu_hotplug(), since they play the role of writers.
- pseries_processor_add() , pseries_processor_remove() now use cpu_maps_update_begin()/cpu_maps_update_done() in place of lock_cpu_hotplug()/unlock_cpu_hotplug().
- Replaced the workqueue_mutex with workqueue_lock, which is a spinlock and guards the workqueues list.
- Updated Documentation/cpu-hotplug.txt to reflect get_online_cpus(), put_online_cpus() in place of the old lock_cpu_hotplug(), unlock_cpu_hotplug().
Patches in v1 No applies recorded yet
| # | Subject and files | Size | Replies | Applied |
|---|---|---|---|---|
| 1/5 | Refcount Based Cpu Hotplug implementationinclude/ init/ kernel/ | +115 | 0 replies | |
| 2/5 | Replace lock_cpu_hotplug() with get_online_cpus()arch/ arch/ arch/ +10 more | +54 | 0 replies | |
| 3/5 | Replace per-subsystem mutexes with get_online_cpus()include/ kernel/ kernel/ +2 more | +38 | 0 replies | |
| 4/5 | Remove CPU_DEAD/CPU_UP_CANCELLED handling from workqueue.ckernel/ | +18 | 6 replies | |
| 5/5 | Update get_online_cpus() in Documentation/cpu-hotplug.cDocumentation/ | +6 | 0 replies |
Reviews trailers per patch
No Reviewed-by, Acked-by or Tested-by trailers for v1 yet.
Cover letter [RFC PATCH 0/5] Refcount based Cpu Hotplug. V2
Hello everyone,
This is the version 2 of the refcount based cpu-hotplug "locking" implementation.
It incorporates the review comments from the first posting which can be found here --> http://lkml.org/lkml/2007/10/16/118.
Changes since v1:
- !CONFIG_HOTPLUG_CPU part is now handled correctly, thanks to the patch from Paul Jackson.
- The cpu_hotplug_begin() uses a waitqueue instead of a completion struct where a writer can wait while there are active readers in the system.
- Provided a new API's cpu_maps_update_begin(), cpu_maps_update_done() for serializing the updates to cpu_present_map and cpu_online_map. Thus threads which update the cpu_present_map should now call cpu_maps_update_begin instead of lock_cpu_hotplug(), since they play the role of writers.
- pseries_processor_add() , pseries_processor_remove() now use cpu_maps_update_begin()/cpu_maps_update_done() in place of lock_cpu_hotplug()/unlock_cpu_hotplug().
- Replaced the workqueue_mutex with workqueue_lock, which is a spinlock and guards the workqueues list.
- Updated Documentation/cpu-hotplug.txt to reflect get_online_cpus(), put_online_cpus() in place of the old lock_cpu_hotplug(), unlock_cpu_hotplug().
I'm Cc'ing the different subsystem maintainers who might be affected by the changes in the patchstack. Especially if they rely on lock_cpu_hotplug() to provide them protection for their local data structures as well.
The patchstack which is based against 2.6.23-mm1 has behaved well when it was stress tested with kernbench running while continuously performing cpu-hotplug operations on i386, x86_64 and ppc64.
Awaiting your feedback.
Thanks and Regards
gautham.
Signature · 5 lines
Gautham R Shenoy Linux Technology Center IBM India. "Freedom comes with a price tag of responsibility, which is still a bargain, because Freedom is priceless!"