[PATCH] smpfatal08: halt secondary RISC-V processors

Gedare Bloom gedare at rtems.org
Mon Apr 12 20:14:22 UTC 2021


On Mon, Apr 12, 2021 at 1:40 PM Jiri Gaisler <jiri at gaisler.se> wrote:
>
>
> On 4/12/21 8:44 PM, Gedare Bloom wrote:
> > Hi Jiri,
> >
> > How much do you think this is a RISC-V specific problem, or one that
> > may affect other SMP processors? Should we add an RTEMS API for this
> > capability instead of shimming some Asm into a test case?
>
> It's bound to happen on any architecture/loader that starts all processor cores before jumping to RTEMS. Leon3 and Griscv only start on core so the test passes, while 'regular' RISCV will fail as all cores are started by the loader. I do not now how the other SMP architectures (x86, PowerPC, ARM) work during start-up. Note that this specific test bypasses most of the normal SMP start-up, that's why the problem occurs. I really don't think we need a special API for this.
>
> I just realized that I can change the patch from an assembly WFI to a simple busy loop instead. This would work on any architecture without assembly. Should I provide a new patch with this cleaner solution?
>
Yes, and a comment in the code please

> Jiri.
>
> >
> > On Sun, Apr 11, 2021 at 1:30 PM Jiri Gaisler <jiri at gaisler.se> wrote:
> >> smpfatal08 fails on SMP RISC-V systems because all cpus are started by the boot-loader and clobber the test output. This patch stops the secondary cpus with a WFI (wait-for-interrupt). Harmless if only one cpu is started by the loader, as in the griscv bsp.
> >>
> >> _______________________________________________
> >> devel mailing list
> >> devel at rtems.org
> >> http://lists.rtems.org/mailman/listinfo/devel
>


More information about the devel mailing list