[LITMUS^RT] Adding Job Deadlines to Control Pages

[Geoffrey Tran]

Hi Björn,

I pulled the new rtspin and it has the same issue:
root at ubuntu:~/liblitmus# ./rtspin -iv 1 100 1 
rtspin/7375:1 @ 0.0229ms
	total interrupts: 0; delta: 0
rtspin/7375:2 @ 100.0650ms
	total interrupts: 0; delta: 0
rtspin/7375:3 @ 200.1190ms
	total interrupts: 0; delta: 0
rtspin/7375:4 @ 300.1220ms
	total interrupts: 0; delta: 0
rtspin/7375:5 @ 400.0690ms
	total interrupts: 0; delta: 0
rtspin/7375:6 @ 500.2010ms
	total interrupts: 0; delta: 0
rtspin/7375:7 @ 600.0860ms
	total interrupts: 0; delta: 0
rtspin/7375:8 @ 700.1200ms
	total interrupts: 0; delta: 0
rtspin/7375:9 @ 800.0870ms
	total interrupts: 0; delta: 0
rtspin/7375:10 @ 900.0969ms
	total interrupts: 0; delta: 0
rtspin/7375:11 @ 1000.1760ms
	total interrupts: 0; delta: 0

Later today, I’ll try reverting my LITMUS install to the current head and see if that fixes it. Will report back here.

Thanks,
Geoffrey

> On Dec 9, 2015, at 8:15 AM, Björn Brandenburg <bbb at mpi-sws.org> wrote:
> 
> 
>> On 08 Dec 2015, at 19:58, Geoffrey Tran <gtran at isi.edu> wrote:
>> 
>> I have been trying to add job deadlines to the control page, per the discussion at https://github.com/LITMUS-RT/liblitmus/pull/1.  However, I seem to be running into a few issues.
>> 
>> To summarize my changes: 
>> - Added lt_t deadline field to control_page struct in rt_param.h
>> - In setup_release() in jobs.c, check for if task t has control page, then write to the field.
>> 
>> In my user space code, I try testing the following:
>> struct control_page* testCtrlPage = get_ctrl_page();
>> if(testCtrlPage!=NULL)
>> {
>>        printf("IRQ: %lu\tdeadline: %llu\tirq_syscall:%lu\n",
>>        	testCtrlPage->irq_count, 
>>        	testCtrlPage->deadline,
>>        	testCtrlPage->ts_syscall_start);
>> }
>> 
>> The code compiles and executes, but the output for all three fields are always 0:
>> “…
>> IRQ: 0  deadline: 0     irq_syscall:0
>> IRQ: 0  deadline: 0     irq_syscall:0
>> …"
>> 
>> I’m not very familiar with this, so there might be something obvious that’s missing.  Any feedback is welcome!
> 
> Hi Geoffrey,
> 
> I can’t spot anything that looks obviously wrong. However, not seeing *any* interrupts in the control paper is rather unlikely; this seems broken.
> 
> I’ve pushed a patch to liblitmus that adds an option to rtspin to make it print interrupt information.
> 
> 	https://github.com/LITMUS-RT/liblitmus/commit/7efbfc0d311523468c3b4665cc31d654c4c58a75
> 
> Running on top of the current master branch, rtspin reports interrupts as expected:
> 
>> [root at litmus-rt ~]# setsched GSN-EDF        
>> [root at litmus-rt ~]# rtspin -v -i   20 100 1 
>> rtspin/1611:1 @ 0.1450ms                    
>>        total interrupts: 0; delta: 0       
>> rtspin/1611:2 @ 100.3330ms                  
>>        total interrupts: 38; delta: 38     
>> rtspin/1611:3 @ 200.3431ms                  
>>        total interrupts: 78; delta: 40     
>> rtspin/1611:4 @ 300.4119ms                  
>>        total interrupts: 118; delta: 40    
>> rtspin/1611:5 @ 400.3479ms                  
>>        total interrupts: 160; delta: 42    
>> rtspin/1611:6 @ 500.3340ms                  
>>        total interrupts: 200; delta: 40    
>> rtspin/1611:7 @ 600.4560ms                  
>>        total interrupts: 238; delta: 38    
>> rtspin/1611:8 @ 700.3901ms                  
>>        total interrupts: 278; delta: 40    
>> rtspin/1611:9 @ 800.3840ms                  
>>        total interrupts: 318; delta: 40    
>> rtspin/1611:10 @ 900.4440ms                 
>>        total interrupts: 358; delta: 40    
>> rtspin/1611:11 @ 1000.4129ms                
>>        total interrupts: 398; delta: 40    
> 
> 
> So I think the control page is working correctly. Looking at your printf statement, is it possible that you are using the wrong bit widths?
> 
> - Björn
> 
> 
> 
> _______________________________________________
> litmus-dev mailing list
> litmus-dev at lists.litmus-rt.org
> https://lists.litmus-rt.org/listinfo/litmus-dev