[LITMUS^RT] ft-trace-overheads

[Björn Brandenburg]

> On 4. Nov 2018, at 14:15, Martinez Garcia Jorge Luis (PS-EC/ESB2) <JorgeLuis.MartinezGarcia at de.bosch.com> wrote:
> I’m running Litmus^RT with a reservation-based plugin on top of a RPi3. While trying to trace and process system overheads by means of the “ft-trace-overheads” script, like this :
>  
> […]
> [ 1153.814976] ftcat invoked oom-killer: gfp_mask=0x24002c2(GFP_KERNEL|__GFP_HIGHMEM|__GFP_NOWARN), nodemask=0, order=0, oom_score_adj=0
> […]

So allocating buffers to hold Feather-Trace samples caused the system to realize that it overcommitted on memory. Essentially, the Feather-Trace buffers were too large for the system to handle. 

> Did you face a similar issue?

I vaguely remember running into similar issues, but don’t recall the specifics. 

> Do you think that vmalloc=512 could help me solve the problem?

I don’t recall off the top of my head what this option does. The first step would be to make sure you are running a minimal system (no unnecessary daemons etc.) to save as much memory as possible. Then try to reduce the Feather-Trace buffer size so that you just barely don’t lose any trace samples. 

Regards,
Björn