Messages in this thread Patch in this message | | | Date | Tue, 21 Feb 2006 10:21:32 -0600 | From | "K.R. Foley" <> | Subject | Re: 2.6.15-rt17 |
| |
Ingo Molnar wrote: > i have released the 2.6.15-rt17 tree, which can be downloaded from the > usual place: > > http://redhat.com/~mingo/realtime-preempt/ > > lots of changes all across the map. There are several bigger changes: > > the biggest change is the new PI code from Esben Nielsen, Thomas > Gleixner and Steven Rostedt. This big rework simplifies and streamlines > the PI code, and fixes a couple of bugs and races: > > - only the top priority waiter on a lock is enqueued into the pi_list > of the task which holds the lock. No more pi list walking in the > boost case. > > - simpler locking rules > > - fast Atomic acquire for the non contended case and atomic release > for non waiter case is fully functional now > > - use task_t references instead of thread_info pointers > > - BKL handling for semaphore style locks changed so that BKL is > dropped before the scheduler is entered and reaquired in the return > path. This solves a possible deadlock situation in the BKL reacquire > path of the scheduler. > > another change is the reworking of the SLAB code: it now closely matches > the upstream SLAB code, and it should now work on NUMA systems too > (untested though). > > the tasklet code was reworked too to be PREEMPT_RT friendly: the new PI > code unearthed a fundamental livelock scenario with PREEMPT_RT, and the > fix was to rework the tasklet code to get rid of the 'retrigger > softirqs' approach. > > other changes: various hrtimers fixes, latency tracer enhancements - and > more. (Robust-futexes are not expected to work in this release.) > > please report any new breakages, and re-report any old breakages as > well. > > to build a 2.6.15-rt17 tree, the following patches should be applied: > > http://kernel.org/pub/linux/kernel/v2.6/linux-2.6.15.tar.bz2 > http://redhat.com/~mingo/realtime-preempt/patch-2.6.15-rt17 > > Ingo > -
Also would you apply the attached patch to fix the RTC_HISTOGRAM Kconfig option. In it's current state "tristate" it allows building as a module, which of course doesn't work.
-- kr --- linux-2.6.15/drivers/char/Kconfig.orig 2006-02-01 12:51:53.000000000 -0600 +++ linux-2.6.15/drivers/char/Kconfig 2006-02-01 12:52:12.000000000 -0600 @@ -712,7 +712,7 @@ module will be called rtc. config RTC_HISTOGRAM - tristate "Real Time Clock Histogram Support" + bool "Real Time Clock Histogram Support" default n depends on RTC ---help--- | |