Trace "timeline" doesn't make sense

[Chris Ergo]

I am working on customizing the DFU bootloader (to add encryption). When attempting to use the instruction trace in Ozone to debug, I see a lot of nonsense. One example is shown below. It is obviously in a loop (evidenced by the repeated cycle of function names), but they show up in the Timeline as nested functions. There seem to be periods of trace missing (shown as "<unknown>"), during which the instruction address jumps around seemingly randomly. Presumably the missing return's are in these areas.

My current hypothesis is that the SoftDevice disables tracing when execution is in it, which has the side effect of instructions being lost for some period of time before and/or after execution is in the SoftDevice.

Is this true? I don't care about tracing within the SoftDevice, but is there any way to get a more complete trace outside the SoftDevice?

[Håkon Alseth]

Hi,

 

My current hypothesis is that the SoftDevice disables tracing when execution is in it, which has the side effect of instructions being lost for some period of time before and/or after execution is in the SoftDevice.

Is this true? Is there any way to get a more complete trace?

The softdevice doesn't disable tracing, its just that your trace debugger does not have the debug symbols to resolve the softdevice functions.

We distribute the softdevice as a binary, so its not possible to debug the functions at the level that you're requesting. All calls to the softdevice address area will come up as "unknown".

 

Kind regards,

Håkon 

[Chris Ergo]

I am not requesting to be able to trace within the SoftDevice. I only want accurate tracing outside the SoftDevice. (I don't care about the trace in the SoftDevice, but it might be a clue to you that the instruction address there jumps around randomly every instruction.)