Re: [PATCH v3 1/3] clk: tegra: Mark HCLK, SCLK and EMC as critical

[Thierry Reding]

On Mon, Mar 12, 2018 at 03:37:28PM +0300, Dmitry Osipenko wrote:
> On 12.03.2018 10:15, Thierry Reding wrote:
> > On Fri, Mar 09, 2018 at 05:35:46PM +0300, Dmitry Osipenko wrote:
> >> On 08.03.2018 17:44, Thierry Reding wrote:
> >>> On Thu, Mar 01, 2018 at 04:33:29PM +0300, Dmitry Osipenko wrote:
> >>>> On 15.01.2018 13:56, Dmitry Osipenko wrote:
> >>>>> On 10.01.2018 16:59, Dmitry Osipenko wrote:
> >>>>>> Machine dies if HCLK, SCLK or EMC is disabled. Hence mark these clocks
> >>>>>> as critical.
> >>>>>>
> >>>>>> Signed-off-by: Dmitry Osipenko <digetx@xxxxxxxxx>
> >>>>>> Acked-by: Peter De Schrijver <pdeschrijver@xxxxxxxxxx>
> >>>>>> ---
> >>>>>>
> >>>>>> Change log:
> >>>>>> v2:     Fixed accidentally missed marking EMC as critical on Tegra30 and
> >>>>>>         Tegra124. Switched to a use of common EMC gate definition on Tegra20
> >>>>>>         and Tegra30.
> >>>>>>
> >>>>>> v3:     Dropped marking PLL_P outputs as critical, because seems they are
> >>>>>>         not so critical. Although, I still haven't got a definitive answer
> >>>>>>         about what exact HW functions are affected by the fixed-clocks.
> >>>>>>         Anyway it should be cleaner to correct the actual drivers.
> >>>>>
> >>>>> Stephen / Michael, would it be possible to schedule these patches for 4.16? My
> >>>>> T20 and T30 devices aren't working without the 'critical clocks' patch. Things
> >>>>> happen to work with the opensource u-boot, but not with the proprietary
> >>>>> bootloader. It's probably not a big deal that out-of-tree devices are broken,
> >>>>> although would be nice to have one problem less.
> >>>>
> >>>> Guys, is there anything I could do to get these patches in linux-next?
> >>>
> >>> I've picked these up into the for-4.17/clk branch in the Tegra tree. I
> >>> already have that branch for the MBIST patches which are a dependency
> >>> for the for-4.17/soc branch.
> >>
> >> Thank you very much! Could you please add stable tag to this ("Mark HCLK, SCLK
> >> and EMC as critical") patch? It would be nice to have 4.16 unbroken eventually
> >> for those who (have to) use downstream android bootloader.
> >>
> >> Cc: <stable@xxxxxxxxxxxxxxx> # v4.16
> > 
> > Should we add a Fixes: line instead? That way it will get automatically
> > backported to all applicable stable releases.
> > 
> > This is a little complicated because the clocks were introduced in
> > different commits, most of them a very long time ago:
> > 
> > 	a7c8485a0ebbdce303c6709e208bb4fd08aff8ad (sclk)
> > 	2db04f16b589c6c96bd07df3f1ef8558bfdb6810 (emc)
> > 	76ebc134d45d7e6e1dc29fdcef4e539c5bc76eb8 (emc)
> > 	...
> > 
> > So Fixes: is perhaps not desirable after all, but then, so isn't tagging
> > v4.16 specifically because this is a bug since almost forever, so v4.16
> > is fairly arbitrary. Shouldn't we at least get this fixed for the last
> > couple of LTS releases?
> 
> That issue was masked in earlier kernels and only appears in 4.16, there is no
> need to backport it further. I thought that adding of 'stable' tag is enough to
> get patch backported automatically, so let's add 'fixes' instead:
> 
> Fixes: 109eba2eb61a ("clk: tegra: Mark APB clock as critical")
> 
> This patch applies cleanly to 4.16, everything should be fine.

It's not immediately obvious why the above would expose the issue and
hence why this patch fixes the problem, and I'm quite frankly not sure
I understand why that is, so perhaps just going with the stable tag is
the best option after all.

Thierry

Attachment: signature.asc
Description: PGP signature