[LITMUS^RT] rebasing on top of Linux 4.1

[Björn Brandenburg]

> On 28 Jul 2015, at 22:51, Björn Brandenburg <bbb at mpi-sws.org> wrote:
> 
>> 
>> On 28 Jul 2015, at 19:38, Björn Brandenburg <bbb at mpi-sws.org <mailto:bbb at mpi-sws.org>> wrote:
>> 
>>> 
>>> On 27 Jul 2015, at 18:00, Björn Brandenburg <bbb at mpi-sws.org <mailto:bbb at mpi-sws.org>> wrote:
>>> 
>>> On 21 Jul 2015, at 19:08, Björn Brandenburg <bbb at mpi-sws.org <mailto:bbb at mpi-sws.org>> wrote:
>>>> 
>>>> 
>>>>> On 20 Jul 2015, at 19:16, Björn Brandenburg <bbb at mpi-sws.org <mailto:bbb at mpi-sws.org>> wrote:
>>>>> 
>>>>> However, there are also some known problem:
>>>>> 
>>>>> 1) When switching to the PFAIR plugin, there’s sometimes a deadlock (+ report from lockdep). Probably some minor fix to adjust to code churn, but this needs to be investigated. 
>>>>> 
>>>>> 2) Budget enforcement is broken: self-suspending tasks + strict budget enforcement causes BUG_ON violations. This also needs further investigation.
>>>> 
>>>> These should both be fixed now. At the moment, the next branch seems to be quite stable in Qemu. Please test with your setups.
>>> 
>>> I’ve pushed a couple more fixes today. The liblitmus test suite has also seen a couple of improvements.
>> 
>> Another bunch of patches. One of them fixes the preemption state machine, which in turn has uncovered a bug in the SRP implementation due to the Linux scheduler changes. The corresponding SRP tests are currently failing.
> 
> Last update for now: I’ve pushed a few more patches that fix the SRP test failure and another race in PFAIR that happened rarely if a task resumed just before the end of a quantum in which it exhausted its current budget. The current version passes all tests and seems to run ok in both Qemu and on my 4-core x86 test machine. Further testing is very welcome, of course.

I’ve pushed another significant update. This bunch includes a number of crash fixes, a fix for the trace infrastructure, and a simplification of the cache-topology relevant parts of LITMUS^RT.

This version seems to be quite stable on my test platforms. I’ll let it run stress tests for a few days, and if nothing else crops up, this will become the next release. I’d appreciate additional test reports, especially on ARM, which I can’t test at the moment.

Thanks,
Björn

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.litmus-rt.org/pipermail/litmus-dev/attachments/20150803/317cb0dd/attachment.html>