Hi James,
On Thu, Nov 10, 2016 at 05:37:22PM +0000, James Hogan wrote:
Hi Fengguang,
On Thu, Nov 10, 2016 at 07:08:43AM +0100, Greg KH wrote:
On Wed, Nov 09, 2016 at 10:00:43PM +0000, James Hogan wrote:
> On Wed, Nov 09, 2016 at 10:22:01PM +0100, Jiri Slaby wrote:
> > On 11/09/2016, 03:46 PM, James Hogan wrote:
> > > Unfortunately the original commit went in to v3.12.65 as commit
> > > 168e5ebbd63e, without fixing up the references to tlb_lo[0/1] to
> > > tlb_lo0/1 which broke the MIPS KVM build, and I didn't twig that I
> > > already had a correct backport outstanding (sorry!). That commit should
> > > be reverted before applying this backport to 3.12.
> >
> > Thanks, reverted and applied. I wonder the builders didn't break given 4
> > mips configurations are tested. I indeed could reproduce locally.
>
> I'm guessing malta_kvm_defconfig isn't one of those defconfigs (and the
> imgtec buildbots don't yet test stable branches). Which builders do you
> use?
I use 0-day for these types of things, and it is not showing up any
errors for the 4.4-stable kernel. Can you get these configurations
added to it so that we can ensure it doesn't regress?
Can we please get a few MIPS defconfigs added to the 0-day testing?
The 0-day build bot should already cover the below configs (not
necessarily in the early hours, but very likely in the first day after
your git push) since they are included in arch/*/configs/. I wonder
where you test your patches? Let me check how they missed the test
coverage.
- malta_kvm_defconfig
this probably doesn't need to be a high priority build, but other
configs don't yet cover MIPS KVM so its worth having (that bit us
recently with 3.12 and 4.4 stable branches).
- 64r6el_defconfig and 32r2_defconfig (4.9 and later)
these are just a couple of the new generic/multiplatform kernel
configurations added in 4.9 (Paul Burton Cc'd). There are others too,
but these will probably give decent coverage. These are likely to be
increasingly relevant as more/new platforms are converted to use it.
(note, the r6 one may require a newish toolchain).
Thanks,
Fengguang
