osmo-trx-lms: Overruns appear from time to time
It was spotted that from time to time osmo-trx-lms tends to not provide data quick enough and some samples are lost. It should be investigated why that happens and find some way to fix this behavior.
Those overruns appear despite rt-prio being enabled (#3971), so let's track this issue here in a separate ticket.
Thanks to fix from #3339, now this kind of overruns are not fatal and osmo-trx sends garbage during those lost samples, but still it's not fine to run into this situation.
This kind of overrun has been seen on high end laptops. It could be related to PM or other processes disturbing the scheduler. It has also been said that it may be related to USB controllers being used.
It could be that we don't see this anymore once we can use rt-prio with LMS (#3971).
#4 Updated by pespin about 1 year ago
There was an announcement in https://github.com/myriadrf/LimeSuite/issues/265 were basically developers from LimeSuite state that "overrun" stat from LimeSuite API during readSamples() was always 0 before some recent change. That means 19.04.1 still contains the bug, and my bet: it was probably fixed in LimeSuite.git 6d34cebbb6f1dad20090c52983924f7c4b0a4acf.
So I'd go for using either master as of today (64bb2b89963ac48706e7990d1f3b60bf65fd5d33), which may provide better feedback on the scenarios related to this ticket.