Johannes Weiner wrote:
>> What's you proposal?
>
> Sounds terrific. I wish you had told us this plan before, that would
> have saved us quite some discussion ;)
>
> So please let us know where to find the merged tree when it's ready.
> We will test it on our S6000 board and look over your work.
I resolved the problems with the merge this evening and posted a
copy here and at kernel.org:
git+ssh://git.linux-xtensa.org/git/kernel/xtensa-2.6.29-smp
git://git.kernel.org/pub/scm/linux/kernel/git/piet/xtensa-2.6.29-smp.git
I have two SMP FPGA systems testing the merge, one with cache aliasing, the other
without. Both have been up now running NTP as a stress test for about two hours
without any problems. Earlier this evening I did have one problem when it was just
sitting idle; looked like a problem I had before and fixed.
I had one problem I've not worked around for now. In vmlinux.lds.in.S you added a
#include of <platform/hardware.h> put this causes compile problems for the LX60
platform. You added it to define KERNELOFFSET where as I was bringing it from
asm/core.h which caused less problems. <platform/hardware.h> seems reasonable,
haven't got a clean solution for the asm macro problem yet for LX60 platform.
Could you try this on your S6000 boards and look over our work. I'll
try using xt-gdb and ocd with the Stretch board. I likely have a few
minor tweaks in the Kconfig for _S6000 for selecting optimal defaults.
-piet