LTTng 0.165 for 2.6.31.5 : refactoring/cleanup
Hi,
I just finished testing the LTTng refactoring. It consists of:
- Change design to inheritance for channels and buffers.
- Eliminate structure duplication for locking/allocation level of buffer and channel.
- Remove now unnecessary krefs.
- Fix cpu hotplug for timers.
- Cleanups.
Impact:
- Cleaner code.
- Simpler structures for channel/buffers.
- Smaller d-cache footprint.
It also permits to change the license of all LTTng modules needed both for kernel and user-space tracing to dual LGPL/GPL license.
Due to the size of this change, review and testing would be very welcome. I decided to do it in a separate patch on top of the tree rather than rebasing. Considering the size of this modification, going over the modified code on a file by file basis rather than looking at the patch itself is probably easier.
I decided to mark the "locked" and "irqoff" locking schemes as "BROKEN" for now and focus on the lockless scheme. We could re-create them if needed by deriving from the lockless code-base.
Thanks,
Mathieu
Signature · 2 lines
Mathieu Desnoyers OpenPGP key fingerprint: 8CD5 52C3 8E3C 4140 715F BA06 3F25 A8FE 3BAE 9A68
LTTng 0.166 for 2.6.31.5 : refactoring/cleanup - 32-bit fix
Hi,
A little word to tell about the updated LTTng 0.166, which fixes a trace session destruction type-size problem on 32-bit machines introduced in 0.165. A simple printk() expecting a size_t rather than an unsigned long caused this.
Mathieu
Quoted 36 lines · * Mathieu Desnoyers ([email protected]) wrote:
Hi,
I just finished testing the LTTng refactoring. It consists of:
- Change design to inheritance for channels and buffers.
- Eliminate structure duplication for locking/allocation level of buffer and channel.
- Remove now unnecessary krefs.
- Fix cpu hotplug for timers.
- Cleanups.Impact:
- Cleaner code.
- Simpler structures for channel/buffers.
- Smaller d-cache footprint.It also permits to change the license of all LTTng modules needed both for kernel and user-space tracing to dual LGPL/GPL license.
Due to the size of this change, review and testing would be very welcome. I decided to do it in a separate patch on top of the tree rather than rebasing. Considering the size of this modification, going over the modified code on a file by file basis rather than looking at the patch itself is probably easier.
I decided to mark the "locked" and "irqoff" locking schemes as "BROKEN" for now and focus on the lockless scheme. We could re-create them if needed by deriving from the lockless code-base.
Thanks,
Mathieu
Signature · 2 lines
Mathieu Desnoyers OpenPGP key fingerprint: 8CD5 52C3 8E3C 4140 715F BA06 3F25 A8FE 3BAE 9A68
Signature · 2 lines
Mathieu Desnoyers OpenPGP key fingerprint: 8CD5 52C3 8E3C 4140 715F BA06 3F25 A8FE 3BAE 9A68