On Thu, Apr 14, 2016 at 10:19:04AM +0200, Geert Uytterhoeven wrote:
> Hi Olof,Simon,
>
> On Wed, Apr 13, 2016 at 10:23 PM, Olof Johansson <olof@xxxxxxxxx> wrote:
> > On Wed, Apr 13, 2016 at 09:57:39PM +0200, Geert Uytterhoeven wrote:
> >> On Wed, Apr 13, 2016 at 9:43 PM, Olof Johansson <olof@xxxxxxxxx> wrote:
> >> > On Mon, Apr 04, 2016 at 10:22:56AM +0900, Simon Horman wrote:
> >> >> Geert Uytterhoeven (2):
> >> >> ARM: shmobile: timer: Fix preset_lpj leading to too short delays
> >> >> ARM: shmobile: timer: Drop support for Cortex A8
> >> >
> >> > I don't understand your naming. THis isn't driver code, this is SoC code.
> >> >
> >> > Anyway, I replied to the drop-A8 patch, the resulting function doesn't make
> >> > much sense once that code path is removed (div is always 1). Seems like this
> >> > should be cleaned up in other ways instead.
> >> >
> >> > So, not merging this branch for now since I think it should be revisited.
> >>
> >> The reason these are small separate patches is because especially the
> >> first one may be backported to stable, as it's a real bug fix.
> >
> > Ok, so then the second shouldn't just remove cortex-a8, but rework the function
> > (yet again).
>
> Sure, reworking it...
>
> Simon: I'm a bit surprised to see that "ARM: shmobile: timer: Fix preset_lpj
> leading to too short delays" got postponed to v4.7, while commit 406663ed61c8
> ("ARM: shmobile: defconfig: Do not enable CONFIG_CPU_BPREDICT_DISABLE") is
> already in v4.6-rc1. The latter makes bad behavior due to too short delays
> more likely to happen.
>
> Sorry for not noticing before.
Likewise, sorry for the mix-up.
I can think of two ways to resolve this, likely there are others:
1. Submit "ARM: shmobile: timer: Fix preset_lpj leading to too short delays"
or conversely
2. Submit a reversion of 406663ed61c8 ("ARM: shmobile: defconfig: Do not
enable CONFIG_CPU_BPREDICT_DISABLE") as a fix for v4.6.
Or perhaps it is a problem that can just be ironed out in v4.7.
Let me know what you think.
