[LITMUS^RT] Feather Trace tools

Björn Brandenburg bbb at mpi-sws.org
Thu Jan 30 09:19:51 CET 2014


On 30 Jan 2014, at 08:43, Glenn Elliott <gelliott at cs.unc.edu> wrote:

> It looks like:
> 1) CONFIG_RELOCATABLE must NOT be set.
> … and that’s it.
> 
> Björn: Is this constraint still necessary?  Didn’t we move to dynamically allocated buffers over the summer?

We did move to dynamically allocated buffers over the summer, but I believe we still need that constraint. Each instance of the ASM trigger code needs to put its address into a table at link time. Here’s the relevant code.

	https://github.com/LITMUS-RT/litmus-rt/blob/master/arch/x86/include/asm/feather_trace_64.h#L30

If it’s relocatable, that address is not known. Making it support relocation is theoretically feasible, but would probalby require serious hacking cycles, which I’m personally not willing to invest as long as !CONFIG_RELOCATABLE is acceptable. If someone knows how to make the events relocatable with little effort, don’t hold back…

Thanks,
Björn

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.litmus-rt.org/pipermail/litmus-dev/attachments/20140130/c8e5aa18/attachment.html>


More information about the litmus-dev mailing list